Archive for the ‘Google Open Source’ Category
The Semantic Robot Vision Challenge (SRVC) is a robot scavenger hunt competition that is designed to push the state of the art in image understanding and automatic acquisition of knowledge from large unstructured databases of images (such as those generally found on the web). In this competition, fully autonomous robots receive a text list of objects that they are to find. They use the web to automatically find image examples of those objects in order to learn visual models. These visual models are then used to identify the objects in the robot’s cameras.
The lastest SRVC was hosted at the International Symposium for Visual Computing (ISVC) in Las Vegas Nevada from Nov 31 to Dec 2, 2009. Five individual teams competed this year and hree of the teams brought robots and participated in both the robot and software league. The other two teams participated only in the software league.
The arena was set up with four chairs, three round tables, two tables with drawers, and a small set of stairs for displaying objects. All of the furniture had at least one object for the robots to discover on it, but not all of the objects in the environment were on the list of items for the robots to find.
The crowd was very interested in watching the different robots moving around the environment during their runs. Unfortunately, the robot teams themselves were plagued with various hardware and software integration troubles and only one team was able to find any objects. However, the robot teams that did not perform well demonstrated that their software was very capable of doing the work in a stand-alone mode. The visual classification results from the software league were very impressive.
The official list of objects consisted of:
- pumpkin
- orange
- red ping pong paddle
- white soccer ball
- laptop
- dinosaur
- bottle
- toy car
- frying pan
- book “I am a Strange Loop” by Douglas Hofstadter
- book “Fugitive from the Cubicle Police”
- book “Photoshop in a Nutshell”
- CD “And Winter Came” by Enya
- CD “The Essential Collection” by Karl Jenkins and Adiemus
- DVD “Hitchhiker’s Guide to the Galaxy” widescreen
- game “Call of Duty 4″ box
- toy Domo-kun
- Lay’s Classic Potato Chips
- Pepperidge Farm Goldfish Baked Snack Crackers
- Pepperidge Farm Milano Distinctive Cookies
Objects 5-9 were part of a list of generic objects that were given in advance to the teams. This was in a response to a suggestion from previous years to allow the teams a chance to try to build classifiers that would be capable of recognizing a generic class of objects rather than a very specific one. You can find a full analysis of the results on the SRVC site.
The US Naval Academy entered a robot based on a iRobot Create platform which used a Hokuyo URG LIDAR for navigation and a camera mounted on a mast for ddetecting the objects. This robot was by far the least expensive of the competitors but was still capable of carrying a laptop as well as the other hardware. However, under this load, the robot rapidly drained its batteries but was still able to capture a few images of objects and label them correctly.
Kansas State University entered with a robot based on a MobileRobots Inc. Pioneer 3 platform. They also had a Hokuyo URG LIDAR for navigation and a camera on a mast used for identifying the objects in the environment. This robot was able to traverse most of the environment successfully. Unfortunately, the robot was not able to aim its camera at enough objects to get a chance to correctly identify them.
The University of British Columbia (UBC) robot had by far the most complex setup of all of the robot competitors. They used a MobileRobots Inc. Powerbot that carried four laptops, multiple cameras–including a monocular PowerShot Canon camera, and a Pt. Grey Bumblebee2 stereo camera, and multiple LIDARs both for navigation and object extraction. The team demonstrated several impressive non-scored runs both before and after the event. However, during their officially scored event, the process that ran the primary object detection camera failed and so they were unable to identify any objects.
For more detailed descriptions of the robots, the software, and the computer vision techniques used by these teams, please refer to the team presentations. Each team’s workshop presentation has been posted to that page. Links to their source code will also be posted.
As this contest continues to grow and evolve, the organizers are quite pleased by the progress of the computer vision research that is being demonstrated at these events. This was shown quite handily by the very high scores in the software-only league. However, the organizers would also like to remind the community that this is a robotics competition and thus want to see advances in active vision techniques, intelligent mapping and exploration, and reasoning about where objects are likely to be found (e.g. the “semantics” of the objects). In previous years, most of the robotics competitors took a random-walk approach to exploring the environment where they would hope to cover all of the space and get enough images to see the objects in question. However, the organizers this year were quite pleased to see the previous reigning champions from the University of British Columbia take the robotic exploration aspect of the competition to the next level. The organizers would like to take the time to highlight some of the significant aspects of the UBC team’s approach to how to control their robot.
The UBC team approached the contest in two distinct phases: a mapping phase, and an object identification phase. The strategy of UBC this year was first to navigate the environment and map it using the SICK LIDAR and a SLAM algorithm (Simultaneous Localization and Mapping). Then the robot would revisit the obstacles in the room and scan them with the Hokuyo LIDAR. Flat horizontal surfaces would be detected in the scans from detection of a few consistent surface normals and a verification stage of the hypothesis of a planar surface. The regions that point out of the plane are interpreted as objects, and 3D bounding are computed from their convex hull (see figure below). This gives a set of candidate locations for the objects. The robot then would revisit these locations to take snapshots and run its object recognition algorithms on these snapshots. Three object recognition methods were implemented, SIFT matching, Contour matching, Deformable Parts Models (DPM). A fourth one using spherical harmonics to recognize 3D data was turned off because it was not quite ready. The DPM approach was trained on the objects known in advance, but could not be used for internet images as it was slightly too slow for that even though it had been rewritten in C.
The organizers were very impressed by the fact that the robot would first identify the specific locations where objects should be found, e.g. the tops of tables and chairs, and then go back and use the 3D sensors to explicitly segment out the locations of the objects to find them. This is exactly the kind of active robotics vision research that we feel will help to push forward the state of the art in real-time computer vision on physical robots and we hope to see more of this kind of approach on future competitors.
To sum up, the research being performed by the teams interested in this competition is extremely impressive. The teams are definitely rising to the challenge put forth by the organizers. Congratulations to all that participated!
By Paul E. Rybski, Robotics Institute, Carnegie Mellon University & Daniel DeMenthon, Johns Hopkins University
The yearly MusicBrainz summit serves an important function in building our community: we talk about issues facing MusicBrainz and we plan the road map for MusicBrainz projects. The summits are usually scheduled to allow as many people to attend as possible and this year we chose Nürnberg, Germany as our location. MusicBrainz contributor Nikolai “Pronik” Prokoschenko lives in Nürnberg and was our local contract and ended up planning most of the summit.
Pronik found us a conference room that we rented for the entire day, complete with open WiFi, which is important if you plan to have a room full of geeks. He also found us a cheap Gasthof that provided lodgings slightly better than a Hostel for a mere 20€ per person per night — a really good deal for Europe. The evening before the summit we all sat in the Gasthof and were treated to some confusing German/Greek cuisine with some of the most rude service any of us have ever encountered. But, our group is used to dealing with the crude Internet public, so we managed to laugh off the horrible service and still have a great time.
To our luck there was a grocery store right next door to our Gasthof and we commenced another successful crowd sourced breakfast. Four people were each given 20€ with the instructions to buy food/drinks that they would like to eat/drink for breakfast/lunch. No collusion was allowed between people! Once the shopping was complete we walked to the conference room, settled in and dove into the masses of food we’d collected. Many tasty bread rolls with jam, nutella, cold cuts and cheese were consumed. Of course we had fun things like a case of Bionade, juices, tea, gummy bears and chocolate. Crowd sourcing breakfast takes a potentially frustrating chore and makes it fun for everyone.
Plus, Pronik and his mate Kira brought a MusicBrainz decorated cake to celebrate 10 years of MusicBrainz!
As people were eating, we started to collect an unconference-like agenda of what people wanted to talk about. We decided to have a detailed state of the project talk including recent developments from meeting our customers in Europe. We also talked about current development processes and some of the problems associated with these processes. Oliver Charles, a 2008 Google Summer of Code™ student, gave an introduction on how to hack on the MusicBrainz server, based on his work from the last year.
Most of the time was spent discussing new features for once we release our much anticipated Next Generation Schema. At times we managed to get into deep philosophical discussions about what MusicBrainz is and what it should be. At other times we discussed light hearted topics with lots of joking. These summits do wonders for building our community and getting people on the same page. We manage to explore many topics and reach consensus on many points in one day instead of spending weeks on the same discussions online.
Finally, in the evening we cleaned up our space and retired to a local beer hall where we continued the discussion in a less formal manner. If you’re interested, we posted all the session notes from the summit on our wiki. All in all, this event was fun and not much effort to put on — thanks to Pronik! On another happy note, 1/3 of the people in attendance were women, which is much better than most tech summits I’ve attended.
In total we spent about $1500, including all the food, drinks, lodgings and one person’s travel costs. For a summit with 12 people, I think we did rather well! I call that Google’s support well spent — thanks again for supporting MusicBrainz, Google!
By Robert Kaye, Executive Director, Metabrainz Foundation
On the 3rd of December we held the latest (and greatest) Google London Open Source Jam at our offices near Victoria. The Jam is a way to get like-minded Open Source contributors and users together and give them a chance to give a 5 minute talk on something dear to their hearts, all the while availing themselves of free beer and pizza!
This time’s topic was the somewhat catchall: “the Web.” Like always, the topic is more of a guide than a rule, so we had some pretty diverse talks.
Our very own Jon Skeet set the evening off to a good start by telling us all about Noda Time — a new Open Source library for handling dates and times in .NET, based on the Joda Time library for Java.
Simon Phillips is a consultant to the film business and gave a great presentation on how he uses Google Wave to help him work closely with directors, script writers, set designers and the like. He showed some great ideas for using Wave in this way and was canvassing for help in developing Open Source Wave robots to help this process.
Simon Stewart gave a rallying cry for making the web more accessible to the blind and deaf, especially in this modern era of HTML canvas and video tags. By ensuring your sites are accessible, you open them up to more users, and as a useful side effect you also make them more testable.
HTTP has started to show its age, and maybe it’s time for a leaner, meaner protocol to come along. I took a brief break from my hosting duties to present a summary of SPDY, a chromium.org project to develop a replacement protocol which will deliver data to our browsers faster.
If you run a web site, you may have come to fear the “Slashdot effect” where you are linked from a popular website and get a spike of traffic. Glyn Wintle from the Open Rights Group (ORG) informed us that this is nothing compared to having a bunch of knitting forums link to you! His was a tale of Open Sourcing of knitting patterns and DMCA take-down notices. He also brought us up to speed on the latest from the ORG.
Sam Mbale gave us an update on his work bringing open source to Africa and told us all about BarCamp Lusaka which he’ll be attending. We look forward to hearing how it went at another Jam.
Robert Rees gave us an experience report on using Velocity templates to divide responsibilities between engineers and web designers. It seems to work pretty well; contracts are enforced by unit tests, and designers know exactly what primitives they can use when laying out web pages.
You can find more pictures of the event on Picasa Web Albums. To find out more about the Google London Open Source Jam, visit http://osjam.appspot.com/. If you’d like to receive regular updates about future jams, sign up for our mailing list. We hope to see you at future jams!
By Matt Godbolt, Software Engineering Team
The Globus Alliance is a community of organizations and individuals developing fundamental technologies behind the “Grid,” which lets people share computing power, databases, instruments, and other on-line tools securely across corporate, institutional, and geographic boundaries without sacrificing local autonomy. We first participated in Google Summer of Code™ in 2008 and we found the experience extremely productive both for the Globus Alliance and the individual mentors, so we wanted to confirm the value of the program for the students who took part. We contacted our eight students from last year to find out what impact Google Summer of Code had on their lives and careers. While many of our students still remembered the experience fondly, and said it was valued highly by prospective employers, there were two students who had particularly remarkable stories.
AliEn Grid Site Dynamic Deployment and Working at CERN
Last year, Artem Harutyunyan, mentored by Tim Freeman, developed a set of scripts on top of Globus Nimbus to dynamically deploy an entire AliEn Grid site (AliEn is the Grid infrastructure which is used by scientists participating in the ALICE experiment at CERN). His collaboration with the CERN and Globus Nimbus folks went beyond his Google Summer of Code work, and resulted in a new framework, called CernVM Co-Pilot, for execution of ‘pilot’ Grid jobs on cloud resources. His work is currently used in production to run Grid jobs from CERN’S ALICE experiment, and there are plans to extend it for the execution of ATLAS and LHCb jobs. Artem also co-authored two papers on his work: “Dynamic AliEn Grid Sites on Nimbus with CernVM” was presented at the 17th International Conference on Computing in High Energy and Nuclear Physics (CHEP 2009) in Prague, and “Building a Volunteer Cloud”, which includes a description of CernVM Co-Pilot, was presented during the Latin American Conference on High Performance Computing in Mérida, Venezuela.
Holder-of-Key Single Sign-On
Joana M. F. Trindade, mentored by Tom Scavo, spent last summer implementing a Holder-of-Key Single Sign-On profile handler for the Shibboleth Identity Provider in Globus GridShib. And, since then, things have just been getting better for her. Thanks to her outstanding summer work, she was offered an appointment as a Visiting Scholar at UIUC, where she worked on researching fault injection in virtual machines with Professor Ravi Iyer. After six months in that position, Joana was offered admission into the masters program at UIUC, where she is currently working with Professor Marianne Winslett. More importantly, Joana tells us that participating in Google Summer of Code gave her a renewed sense of confidence in her research abilities, having previously thought that her academic background was insufficient to gain admission into a top-tier university in the US. Joana tells us that “After Google Summer of Code, I regained that hope, and I must say I’m really happy to have found a topic in Globus to which I could contribute, and that in turn opened so many doors for me.”
Congratulations Artem and Joana for all you have achieved!
Lessons Learned
Our first Google Summer of Code last year also had its fair share of challenges, including two students who didn’t make it through the program, but it gave us the opportunity to learn a lot about how to mentor and manage summer students. We were fortunate to be selected again this year as a Google Summer of Code mentoring organization, which allowed us to apply everything we learned. First of all, we required students to provide more information about their background and the project they were proposing. Last year our student application form was essentially a blank form saying “Tell us about your project here,” so this year we presented prospective students with more specific questions. We also decided to check in with our students more often which, at least in one case, allowed us to identify a problem between a student and a mentor early on, giving us time to deal with it constructively before the midterm.
In the end, applying what we learned during last year’s Google Summer of Code and as well as the Mentor Summit had a noticeable effect. We were fortunate to be given ten students to mentor, and all ten students passed. Furthermore, our mentors report that practically all the code written by the students has either already been released or will be released soon. In fact, overall, we felt that this year’s students rocked. Here’s a summary of their summer work.
Going Beyond a Single Cluster
The Globus Nimbus cloud toolkit allows you to turn your cluster into an Infrastructure-as-a-Service (IaaS) cloud. However, it was mainly geared towards managing a single cluster. Not any more! Adam Bishop, mentored by Ian Gable, worked hard over the summer to add new components enabling multiple cluster support for Nimbus. He developed a series of production-quality plugins, which have already been committed to the Nimbus source repository, that publish the state of Nimbus cluster back to a Globus MDS Registry. This allows the availability of cloud resources across multiple Nimbus clusters to be gathered together into a single registry, which is the first step towards adding cross-cluster support to Nimbus.
Spilling Over Multiple Clusters
Another student, Jan-Philip Gehrcke, mentored by Kate Keahey, also spent the summer with his head in the clouds, but in a good way: he developed the Clobi project, a job scheduling system supporting virtual machines (VMs) in multiple IaaS clouds, with support for Globus Nimbus and Amazon EC2 clouds. In a nutshell, there are many scientific applications that are typically run as “jobs” on a compute cluster. Jan-Philip’s project allows these jobs to be submitted to a cloud instead of to a traditional compute cluster. The most interesting use case is when a site operates a Globus Nimbus cloud and, during peaks in demand for computational capacity, extends its capacity momentarily by spilling the jobs over to a second (or third, or fourth, …) cloud such as Amazon EC2. Although Clobi is not tied to any particular application (its design is generic and should be useful whenever it’s convenient to distribute jobs across different clouds), the motivating application for Clobi is ATLAS Computing (for the LHC‘s ATLAS experiment at CERN). In fact, by the end of the summer, Jan-Philip was able to run a common ATLAS Computing application (the so-called “full chain”) successfully with Clobi. If you want more details about Clobi, check out this blog post written by Jan-Philip.
Incremental GridFTP Transfers
Enough about clouds, let’s move on to the exciting topic of data. Globus GridFTP is a high-performance, secure, reliable data transfer protocol that is pretty good at moving data. Fast. Of course, there’s always someone who wants to go even faster, like Shruti Jain, mentored by Michael Link. Shruti took globus-url-copy, the GridFTP client, and added a ‘sync’ feature that allows a local and remote file to be synchronized, by sending only the changed sections of the file. This results in more effective bandwidth utilization by avoiding redundant data transfers.
Checksummed GridFTP Transfers
Remember Mattias Lidman? We certainly do. In last year’s Google Summer of Code, he developed a compression driver for the Globus XIO input/output library (which GridFTP depends on) to compress/uncompress data as it passes through it. However, although moving data faster is all good and well, it’s not worth much if it somehow gets corrupted in-flight. So this year, Mattias, mentored by Joseph Bester, continued to work on Globus XIO and developed a Checksum Driver. Mattias’s driver checksums GridFTP data streams allowing both ends of a GridFTP transfer to verify the integrity of the data.
CQL Queries Builder
You know one really cool thing grids are used for? Cancer research. The Cancer Biomedical Informatics Grid, or caBIG®, is an information network enabling all constituencies in the cancer community – researchers, physicians, and patients – to share data and knowledge. caGrid is the underlying service-oriented infrastructure that supports caBIG, and it relies heavily on the Globus Toolkit. Some of the data services in this architecture use a query language called CQL that is, well… complicated. To make life easier for scientists, Monika Machunik, mentored by Wei Tan, wrote a plug-in for Taverna (an open source tool used by scientists to design and execute workflows) for constructing CQL queries, allowing scientists to focus on their work rather than on the intricacies of the CQL language.
GridWay-Google Maps Mashup
Grids require coordinating resources across multiple organizations, and the Globus GridWay meta-scheduler is a great tool to do just that. However, coordinating hundreds or even thousands of machines across dozens of sites can get a bit messy using the console-based tools included with GridWay. Carlos Martín, mentored by Alejandro Lorca, tackled this problem by creating an interactive GridWay-Google Maps mashup, allowing the administrators and users of a GridWay installation to get a quick snapshot of the status of multiple sites and the jobs running in them, as shown in this screenshot:
Carlos used the Google Web Toolkit to develop this application, which is totally decoupled from GridWay, making it easy to install it alongside existing installations of GridWay. In fact, you can download the GridWay+Google Maps application and check out its documentation, including more screenshots, at the application’s page on the GridWay site.
GridWay GUI
Srinivasan Natarajan, mentored by Jose Luis Vazquez-Poletti, worked on a more administration-oriented GUI for GridWay, allowing users to compose, manage and control their jobs instead of using the command line interface. This GUI includes a host of other features, such as host and user monitoring, filtering account statistics and execution history information, and support for processing DAGMan workflows, including visualizing dependencies between jobs in the workflow.
Both of the GridWay projects were presented in several sessions, including one on nuclear fusion, at the EGEE’09 conference in Barcelona, Spain back in September.
GridFTP Benchmarking
How about we get back to the subject of data management? The recent addition of UDT (UDP Data Transfer) support to GridFTP has made even faster transfer speeds possible. You guessed it: here’s another student who couldn’t resist the need for speed this summer. Jamie Schwettmann, mentored by Raj Kettimuthu, sought to characterize the performance of GridFTP over 10Gb/s networks, specifically to measure the speed increase given by UDT as compared to TCP transfers, as well as a number of other considerations such as CPU and memory overhead at both ends of the transfer. In doing so, they decided to develop an automated GridFTP benchmarking and throughput optimization utility called globus-transfer-test, which takes URL pairs from a list or on the command line, and allows for varying input parameters such as parallelism level, transfer type (memory-to-memory, disk-to-disk, etc), TCP Buffer Sizes, MTU sizes, and all other standard globus-url-copy options (except multicasting) and when possible, compares with other performance and throughput utilities such as iperf or scp. Designed for general use by users or administrators as well as to carry out our performance characterization, globus-transfer-test aims to provide enough information to optimize GridFTP options for maximizing throughput between grid sites. This common need has allowed collaboration with many other projects and organizations in the course of development and testing, including the US ATLAS Project, TeraGrid, and OSCER. Jamie even presented a poster on her project at the 2009 Oklahoma Supercomputing Symposium.
AJAX Framework for Globus Web Services
Many of the components in Globus are web services, which are not exactly human-readable creatures. Fugang Wang, mentored by Tom Howe, developed a JavaScript API that enables accessing Globus services from a web client using AJAX. Fugang’s framework, which includes a backend service that mediates service requests to the Globus toolkit and an AJAX web client to access this services, makes life easier for Globus developers and users by allowing them to interact with Globus services from the comfort of their web browsers.
Secure Cloud Communications
And we’ll end with the ever-popular subject of data management. Melissa Weaver, mentored by John Bresnahan, developed a PSK driver for Globus XIO. She first developed a program that, using OpenSSL libraries to encrypt and decrypt data using a stream or block cipher of the user’s choice, allowed her to experiment with different lengths of keys and initialization vectors and different file sizes to make performance measurements. Then, she developed the XIO PSK driver itself, which used the results of the first program to implement an RC2 block cipher to ensure any communication between computers, once a connection has been set up, is secure.
High energy physics experiments at CERN! Cancer research! Nuclear fusion! Cloud computing! Fast data transfers! Oh my! Oodles of congratulations to our mentors and students for all their hard work and for making this such an awesome Google Summer of Code for the Globus Alliance!
By Borja Sotomayor, Ph.D. Candidate, University of Chicago and Google Summer of Code Organization Administrator
Slow DNS servers can make for a terrible web browsing experience, but knowing which one to use isn’t easy. namebench is a new open source tool that helps to take the guess-work out of the DNS server selection process. namebench benchmarks available DNS services and provides a personalized comparison to show you which name servers perform the best. As a System Administrator at Google, I was curious about measuring how BGP route selection affected the performance of Google Public DNS. This curiosity resulted in writing a small benchmarking script, which was further developed during my 20% time to become a full-featured application for Windows, Linux, and Mac OS X.
namebench is covered by the Apache 2.0 license, and was made possible by using several other great open-source tools including Python, Tkinter, PyObjC, dnspython, jinja2 and graphy. It also makes use of the Google Chart API to visualize the results:
In order to provide the most relevant results, namebench employs a number of interesting techniques. First, it personalizes the benchmark by making use of your browser history to see what hosts to benchmark with. It also determines cache-sharing relationships between different IP’s and removes the slowest of these servers to avoid improperly benchmarking them solely on cached results. namebench will also report on DNS misbehavior such as DNS hijacking and censorship.
namebench 1.0 is available for download now. If you would like to discuss or have any questions namebench, please join the namebench mailing list. Happy hacking!
By Thomas Strömberg, Hardware Operations Team
The Joomla! project was thrilled to sponsor 18 Google Summer of Code students for 2009, and we are pleased to report that 16 (89%) successfully completed their projects. Most of the projects were based on ideas generated by the Joomla! community, and our community seems to be very excited about the results.
Our two primary goals for Google Summer of Code 2009 are to (1) develop relationships with student developers that will encourage them to continue working in the project; and (2) add features and functionality to the Joomla! CMS. Our participation in Google Summer of Code 2009 was very successful on both fronts.
Relationship to the Project
Several of our students this year were already contributing to Joomla! prior to participating in the program, and the Google Summer of Code experience has only strengthened that relationship. For example, one of our students, in addition to completing his project, is now a leader in the release of the next Joomla! version. At least two students (so far) have officially joined project working groups, and several others have contributed to the project over and above their Google Summer of Code projects. Many other students have also expressed interest in continuing the development of their code beyond the program timeframe.
This year, at the end of the term, we gave each student the opportunity to present a webinar where they could demonstrate their project to the community. Even though it was a lot of extra work, more than half the students did this. The results were excellent, and the students did really good, concise, focused presentations. We recorded and linked to the webinars on our site so that anyone in the community who is interested in the Google Summer of Code work can simply watch a short webinar to see an actual demonstration of the projects.
Using the Code
There are three ways the code from Google Summer of Code projects can be used within the Joomla! CMS. In some cases, some or all of the code will be incorporated directly into the core codebase for the upcoming Joomla! version 1.6. In other cases, the code has been published as an extension that can be downloaded and used by any Joomla! user on their website. The third method is that the code will be used as a basis for further work.
Some students have combined two of the methods above, for example, producing an extension for the current version 1.5 and making the code available for the core in our version 1.6.
More Information
We invite you to visit our Joomla! Community site for more information about the different projects and what was accomplished, and to download the code.
By Mark Dexter, Joomla! Project
The Etherboot Project is very pleased to have participated in Google Summer of Code™ 2009. This summer marks our fourth consecutive annual participation in this excellent mentoring program.
Google generously sponsored five students to work with us, and four of our five students (80%) successfully completed their projects. We would like to thank Google, our mentors, and our students for making this a pleasant, productive, and memorable summer.
We particularly wish thank one of our mentors, Stefan Hajnoczi, who was a GSoC student with us last year. His insights and diligence are extremely helpful and enlightening.
Although all of our GSoC projects were not successfully completed, our students’ work was generally of excellent quality, and we sincerely thank them all for their diligence. Our participation in GSoC has strengthened our project by encouraging us to create additional technical and social infrastructure. These improved facilities make it easier for new people to become involved with our project and also help us better support and communicate with our existing community.
We look forward to giving future GSoC students and other interested and motivated people a positive introduction to FOSS development. What follows is a brief summary of our 2009 students’ work with links to their full project pages. We conclude with a brief outline of our mentoring system that we hope may be helpful to other projects.
Student Project Summaries
Daniel Verkamp
Daniel implemented an automated regression testing framework to help us consistently deliver high-quality releases.
Joshua Oreman
Joshua extended gPXE, our network bootloader, with an 802.11 wireless stack, and added drivers for two wireless cards.
Lynus Vaz
Lynus extended gPXE scripting with a more powerful language that is capable of expressing advanced boot policies.
Pravin Shinde
Pravin created a central resource to network boot operating systems, diagnostic tools, and utilities at http://boot.kernel.org/.
Chris Kluka
Chris worked on adding a network driver DLink DGE-530T ethernet cards. Though unable to complete his project, he compiled and created useful information which will facilitate future work on this driver.
Our Mentoring System
Over our years of GSoC participation we have developed and refined a system for mentoring that works quite well for us. One of the most important attributes of our system is that we break the twelve week GSoC coding period into twelve one week evaluation periods. By doing this we ensure that we always have recent information on how each of our students is doing, which allows us to intervene in a timely fashion when needed.
Techniques
Here are some of the other ways we structure our GSoC participation:
* We mentor as a team. We have a mailing list and private IRC channel specifically for mentors.
-Project Plan
-Journal (broken into twelve weeks)
-Notes
-git repository link
By Marty Connor, Etherboot Project
The Apertium Project works on open-source machine translation and language technology. We try to focus our efforts on lesser-resourced and marginalized languages, but also work with larger languages. To date, we have released translators for 21 language pairs, covering languages spoken by 1.1 billion people, ranging from English (est. 500m speakers) to Aranese (est. 4,000 speakers). A similar number of additional language pairs are in development. The Apertium software is licensed under the GPL, but in addition (a rarer situation in the machine translation field) so is the data for all these language pairs. This means that the data can be re-used by other language projects (e.g. in developing spelling or grammar checkers, thesauri, etc).
This was our first year in Google Summer of Code and we were very fortunate to receive nine student slots. We filled them with some great students and are pleased to report that out of the nine projects, eight were successful.
The completed project were:
A translator for Norwegian Bokmål (nb) and Norwegian Nynorsk (nn)
This project was accepted as part of our “adopt a language pair” idea from our ideas page. Some work had already been done on the translator but it was a long way from finished. Kevin Unhammer from the University of Bergen was mentored by Trond Trosterud from the University of Tromsø. The final result, after an epic effort, is a working translator (and the first free software translator for nb-nn) that makes a mistake in only 11 words out of every 100 translated, making using the system for post-edition feasible.
One of the key aspects of Kevin’s work was the re-use and adaptation of existing open source resources. Much of the bilingual dictionary was statistically inferred from the existing translations in KDE, using ReTraTos and GIZA++ (created by Franz Och). In addition to this, Kevin used the Oslo-Bergen Constraint Grammer, contributing fixes not only to that, but to the VISL CG3 software itself. After the GSoC deadline, Kevin has continued his work, including incorporating some changes from feedback from the Nynorsk Wikipedia.
A translator for Swedish (sv) to Danish (da)
Another language pair adoption, Michael Kristensen, who had previously done some work on this translator, was mentored by Jacob Nordfalk, the author of our English to Esperanto translator. As there are very few free linguistic resources for Swedish and Danish the work was pretty much started from scratch, although we took great advantage of the Swedish Wiktionary. The translator is only unidirectional, from Swedish to Danish, and it has an error rate of around 20%.
The completion of this translator is something of a triumph for Apertium. Begun back in 2005, the project had been neglected for many years. This was the first translator for the Apertium platform that focused on non-Romance languages.
Multi-engine machine translation (MEMT)
Gabriel Synnaeve was mentored by Francis Tyers to work on a module to improve the quality of machine translation by taking translations from different systems and merging their strengths and discarding their weaknesses. The two systems focused on in the initial prototype are Apertium (rule-based MT) and Moses (statistical MT) but it can easily be extended to more. The idea behind the system is that for some languages there is often not one MT system which is better than all others, but some are better at some phrases and some are better at others. Thus, if we can combine the output of two or more systems with different strengths/weaknesses, we can make better translations.
Perhaps the most exciting aspect of the MEMT project is its potential for use as a research platform for future work on hybrid machine translation, by allowing the researcher to focus only on the algorithms they wish to implement. During the project, Gabriel was joined by Francis in person for a ‘mini-hackathon’, which, despite something of a farcical start involving requests made on IRC for phone calls across Europe on behalf of two people who were in the same city, lead to a greater degree of functionality and modularization in the code.
Highly scalable web service architecture for Apertium
Víctor Manuel Sánchez Cartagena worked with mentor Juan Antonio Perez-Ortiz on a highly-scalable web service architecture, or, Apertium for Cloud computing. Initially targeting Amazon’s EC2, as well as standalone servers, the scalable web service allows the use of multiple translation services on multiple physical or virtual servers, scaling to meet the translation demands of users, from a single user-facing service, which implements the Google Language API.
The core of the system is the translation router, which controls the flow between user and translation server, based on a variety of factors, including the availability of the language pair, the current load on the server, as well as providing a framework to allow these factors to have different priorities on a per-user basis. It also takes into account the cost of each translation request. The project is a complete package; as well as the router, it includes a translation daemon, and convenience scripts to ease the rollout of server instances.
In addition to his work on his project, Víctor is also serving as an organiser for the FreeRBMT workshop.
Conversion of Anubadok
Abu Zaher was mentored by Kevin Donnelly and Francis Tyers to convert Anubadok, an open-source MT system for English to Bengali to work with the Apertium engine. This was an ambitious project and not all of the goals were realised, but we were able to make the first wide-coverage morphological analyser / generator for Bengali and a substantial amount of lexical transfer, so the project was a great success.
Zaher is also looking at improving the Ankur spell checker with information from his analyser / generator, so the work done is already being reused; there is also interest in using the data to create a Bengali stemmer, for more efficient searching/indexing of Bengali texts, and a number of tools which were created to model the various aspects of Bengali inflection will certainly prove useful in other areas of NLP for Bengali.
Apertium going SOA
Pasquale Minervini’s work was motivated by the needs of Informatici senza Frontiere to have a translation engine that would fit into a Service-Oriented architecture. To this end, Pasquale, mentored by Jimmy O’Regan, designed an XML-RPC-based server that efficiently contains the Apertium pipeline, and layered it with JSON (still under development), SOAP, and CORBA services, which, as well as making Apertium more buzzword compliant, gives a greater range of options to programmers wishing to integrate Apertiums translation services into a wider range of architectures. This is undoubtedly a popular project idea: Alexa’s keywords for Apertium show ‘apertium going soa’ and ‘deadbeef apertium’ (deadbeef is Pasquale’s IRC nick) in 2nd and 4th place for search keywords leading to Apertium.
Because of the potential overlap between their projects, in the first weeks of their GSoC work, Pasquale and Víctor agreed on the Google Language API as a standard for their projects to communicate; Pasquale took this agreement one step further by implementing the ‘language detection’ feature of the API – something previously unavailable in Apertium. In addition to that, Pasquale also contributed memory leak checks against the Apertium platform, as well as other fixes, and has helped another (non-GSoC) student in the goal of porting Apertium to Windows.
Trigram part-of-speech tagging
Zaid Md. Abdul Wahab Sheikh was mentored by Felipe Sánchez Martínez to improve our part-of-speech tagging module to use trigrams instead of bigrams, as well as implementing changes to the training tools to create data for it.
Apertium was originally designed for closely related languages, but is growing to meet the challenges of translating between more distant languages. One of the unique aspects of Dr. Sanchez’s work on Part-of-Speech tagging is the use of target language information which allows an accurate tagger to be trained using much less data than usual. Zaid’s work builds on Dr. Sanchez’s work with first-order Hidden Markov Models, extending it to second-order HMMs, similarly to TnT. This enables more accurate translation between more distant languages, using the same methods, so that the rest of the Apertium system can continue to grow.
Java port of lttoolbox
Raphaël Laurent worked with Sergio Ortiz Rojas to port lttoolbox to Java. lttoolbox is the core component of the Apertium system; as well as providing morphological analysis and generation, it also provides pattern matching and dictionary lookup to the rest of Apertium, so a Java port is the first step towards a version of Apertium for Java-based devices. Raphaël finished an earlier line-for-line port contributed by Nic Cotrell, first making it work; then making it binary compatible.
As it stands currently, lttoolbox-java can be integrated into other Java-based tools, facilitating the re-use of our software and our extensive repository of morphological analysers. Tools such as LanguageTool, the open source proofreading tool, also make extensive use of morphological analysis, but OmegaT, the open source CAT tool, could use it for dictionary look-up of inflected words; it could even be used with our own apertium-morph tool: a plugin for Lucene that allows linguistically-rich document indexing.
FreeRBMT
On the 2nd and 3rd of November, we held the first FreeRBMT workshop, which was heavily inspired by the Google Summer of Code program, both as a way for students and mentors to meet in person, and to provide the students with an opportunity to present peer-reviewed papers about the work they completed during the program. The entire proceedings are available from the University of Alicante; in particular, we would like to highlight the papers which were successfully presented by the students who took part in GSoC:
Apertium goes SOA: an efficient and scalable service based on the Apertium rule-based machine translation platform; Minervini, Pasquale
Development of a morphological analyser for Bengali; Faridee, Abu Zaher Md.; Tyers, Francis M.
An open-source highly scalable web service architecture for the Apertium machine translation engine; Sánchez-Cartagena, Víctor M.; Pérez-Ortiz, Juan Antonio
Reuse of free resources in machine translation between Nynorsk and Bokmål; Unhammer, Kevin; Trosterud, Trond
A trigram part-of-speech tagger for the Apertium free/open-source machine translation platform; Sheikh, Zaid Md Abdul Wahab; Sánchez-Martínez, Felipe
In addition, the following paper was presented by the mentors of a successful project (Michael, the student, was unfortunately too busy to participate in its writing):
Shallow-transfer rule-based machine translation for Swedish to Danish; Tyers, Francis M.; Nordfalk, Jacob
We would like to thank Google for providing us with the opportunity to participate in the Summer of Code program; in particular, Leslie, Cat, and Ellen, for making it run so smoothly. We would also like to make special mention of two students: Ankitha Rao and Daniel Beck, who, despite being unsuccessful in their applications, continued to work on their proposed projects (an English to Hindi translator, and a module for multi-word units, respectively). Finally, we would like to thank all of the students, mentors, and administrators who contributed their time and skill to Apertium.
By Francis Tyers and Jimmy O’Regan, Summer of Code Mentors for the Apertium Project
SWIG is a programmer’s tool designed to make it easier to use C and C++ code from other popular programming languages such as Python, Perl, Ruby, PHP, Java, and C#. 2009 was SWIG’s second Summer of Code, and this year we mentored five projects related to SWIG. All five students were very active over the summer period and produced some great new features. In no particular order:
Matevž Jekovec has been busy working at the coal face of SWIG to add support for C++0x, the forthcoming C++ standard. Matevž has managed to achieve close to full support for C++0x. The C++0x Wikipaedia article details the numerous planned new features and Matevž has put together a SWIG C++0x page documenting the new SWIG support for each of these. In summary the enhanced C++ language can now be parsed by SWIG, which in itself is a great step. There is much more than just this though, as most of the information parsed is used to create useful wrappers of C++0x code. The work can be tried out on the C++0x branch which should be merged fairly soon into a forthcoming release.
Miklos Vajna has been working on SWIG’s PHP support to implement an advanced SWIG feature already supported for most other target languages, but not PHP. The feature is called “directors” and allows cross-language polymorphism – wrapped C++ classes can be subclassed in PHP and virtual method calls work in the natural way, whether they’re made from PHP or C++ code. You can read more in the new PHP Director documentation. Miklos made such great progress that we were able to merge this support into SWIG 1.3.40, which was released even before the Summer of Code finished. Miklos also spent some time working on improving SWIG’s test suite for PHP, and fixing bugs in the PHP support.
Ashish Sharma spent the summer adding support for Objective-C as a new target language. Objective-C is a major language on the Mac OS X platform. This means that now SWIG can be used to generate Objective-C wrappers over C++ code. In particular the wrappers include proxy classes, which preserve the class hierarchy from the C++ code. Ultimately this means that from the user’s perspective, proxy objects look no different to objects originally written in Objective-C. Adding a new target language is quite a considerable task and Ashish is keen to add plenty more improvements over the coming months. Ashish’s work is in Subversion and can be accessed in the ashishs99 branch.
Baozeng Ding has also added a new target language, in this case for the Scilab language, a free numerical computing package. He has coded up support for all the C features: variables, functions, constants, enums, structs, unions, pointers and arrays and also intends to develop it further in the near future. Documentation for SWIG and Scilab can be viewed online direct from Baozeng’s Subversion branch.
Kosei Moriyama has been working on Perl bindings for the Xapian library using SWIG, to replace some existing bindings implemented by hand. He’s achieved almost complete compatibility with the API of the existing bindings (the only real omission is callbacks which are waiting for completion of director support for Perl in SWIG). He has also wrapped features which weren’t previously accessible from Perl. You can view Kosei’s work online in his Subversion branch.
Finally, many thanks to Google for sponsoring the Summer of Code and a special thanks for all the hard work done by the students, mentors and Olly Betts, the co-administrator.
By William Fulton, SWIG administrator
In July we announced that we were working on a project called Google Chrome OS, an open source operating system based on the Google Chrome browser and built for today’s web. For the past few months we have been working hard on developing a solid foundation and today we are excited to announce the Chromium OS open source project.
You can read more about our open source announcement at the Chromium Blog, or get involved directly at chromium.org. We look forward to working with the open source community to help shape the future of personal computing.
By Martin Bligh, Software Engineer
