SECOND MEETING OF THE OPEN-SOURCE SPEECH RECOGNITION INITIATIVE, INC.
("OSSRI")
Thursday, 26 February 2004
Location: OSSRI listserve
and 312 Washington St 4-1, Wellesley MA 02481 USA
The meeting came to order late, at 07:49:12 -0500, or 2:49 GMT.
ATTENDING
DIRECTORS:
Susan Cragin
John Dowding
Jessica Hekman
Volker Kuhlmann
Tristram Metcalfe
Edwin Suominen
Ivan Uemlianin
Dustin Wish
MEMBERS:
Mark Kyes
Turner Rentz III
GUESTS:
None.
OPENING REMARKS:
SUSAN. Our president, Dustin, is incapacitated and I have taken over running the meeting.
DUSTIN. Thank you for calling the meeting to order. I have over 101 degree fever and have been in bed since yesterday morning. I just got up to check my email.
IVAN. First of all, thanks to Susan for taking over the meeting at such short notice!
ED. Note to Dustin: sorry I didn't get you the info. It can wait for the next meeting. Have been fighting recurrent pneumonia. I think you, of all people, will understand right now!
AGENDA:
1. NOMINATION OF MEMBERS.
1.) Saravana Krishnamurthy, Manager-Engineering, SS8 Networks, Inc. from San Jose, CA, USA
2.) Christophe Gerard, NMS Communications, Shrewsbury, MA, USA
3.) Alfred Dayton, Bay Shore, NY
4.) Turner Rentz III
5.) Dennis A. Beckley
6.) Bruce Cyr
VOTE: CONSENSUS
DISCUSSION:
IVAN. Welcome, new members! It might be a nice idea for newcomers to introduce themselves to the list (see point 4 below). That is, unless we want to feel like a big mutlinational corporation.
TURNER. This looks like a well rounded council. I am focussing on Accent recognition.
ED. We are fortunate to have their interest and good will.
VOLKER. Do we really need to vote on each new member, isn't the OSSRI open to anyone?
SUSAN -- NOTE. Perhaps in future, new members should be a non-voting issue.
2. PURCHASE AND HOSTING FOR OSSRI.org
I (Dustin) have personally volunteered to provide hosting for the site. We have a member that volunteered to purchase the domain name for OSSRI. I would like to see the membership discuss this before the meeting so we may get this issue to vote during the meeting.
VOTE: YES-8; NO-0
DISCUSSION:
MARK. Both of these offerings sound great. for the domain, I think the Adminstrative, technical and billing contacts can be assigned to roles that we can define people to fill those roles as needed later..
TURNER. The word "opensource" will soon become one word in the parlance, and the dns osri.org is easier to remember and type. "ossri" sounds like you're saying "sorri" >:)
IVAN. YES to ossri.org (not in favour of the osri.org suggestion)
3. PLAN OF ACTION.
We should have a preliminary plan of action for the technical side. Discussion? Thoughts?
JESSICA. Though I haven't done any work on them lately, I am affiliated with the xvoice and xvoice-sphinx projects. I would love to have people take a look at xvoice-sphinx.sourceforge.net and see if that's a good base for us to start with. xvoice-sphinx was researching getting a pre-existing open source engine to reliably recognize dictated speech. As of the time we gave up, we were unable to get Sphinx to perform reliably. It might be worth a) investigating Sphinx more -- more details on that web page -- or b) looking into other engines.
DUSTIN. Once the site is setup we can link to all major projects and start a CVS code area for the start of our own. The next step is the programming....
MARK. Definately, I think it prudent to come up with a clearly defined set of goals, once the goals have been defined, we should form teams that are responsible for attaining those goals, these teams wouold evaluate technologies, define architecture and craft designs for the goal(s) they've been assigned (just enough so that the larger tasks can be broken off if need be and distributed to others to work on). It's my belief that our process should be as nimble and as light as possible so that we don't get bogged down in to much process but can keep the project moving and our codebase growing. Finally, we'll need a person (maybe more).. that looks at the sytem from a higher level this person/group ensures that the architectures being proposed by the teams are compatible with each other. he/she/they spread good ideas from group to group and becomes the resource the teams go to when they disagree. Finally, I think the teams should strive to come up with a very high level plan of what capability they are looking to have and when it will be available, if for no other reason than they have something to track to and other groups will know when capability will be available for them to use..
IVAN. As Jessica and Mark have suggested, I think a good early goal would be to evaluate the technologies currently available, upon which an open source ASR system could be based. This would mean doing some kind of comparative assessment of Sphinx, ISIP, JULIUS, and no doubt there are others.
One difficulty in carrying out such an assessment might be the lack of training data and/or acoustic models, so we might need to trawl around to see what training data is available at a reasonable price (e.g. free). The above would teherefore inculde evaluation of acoustic model and language model building tools.
It might be worth considering what is the minimal useful functionality for an ASR application: we might come up with something achievable, even if it's only a demo. I'm thinking along the lines of a plug-in for Mozilla or OpenOffice.org - would it be useful to have something that could recognise the names in your address book? or something that would allow a user to navigate GUI menus by speech? (This might mean talking to some end-users;)
Finally, I am no legal authority (slight understatement), but I have a hunch that 'interesting' IP issues may crop up - beyond the 'GPL or BSD?'. For example, some material (e.g., training data from ELDA; various corpora and/or pronunciation dictionaries) is available either only for, or more cheaply for, non-commercial use: if we used this material to build language or acoustic models would that constrain who could use (or whether we could sell) the models (or application using them)? This question may just demonstrate my ignorance of the law, of course, but it does seem 'these days' that IP law is a fairly important component in open source projects. At the very least I'm interested in earning more about this area.
So, some ideas:
(a) evaluate available ASR software
(b) source available training data (for acoustic and language models)
(c) build a library of minimal use cases
(d) build a library of IP issues and their implications
TURNER. The group should focus, in my view, on the technical aspect by focusing on three things. specification, specification, and the specification.
A good plan going forward from that, is to divide up the group into XP units and deal out specific parts based on the dilbert principle.
4. SUBSCRIBERS
We now have 40 subscribers to the listserv. Several contacted Susan off-list to ask how they can help. Should she refer them to all the directors, or would they prefer that one person be designated?
DISCUSSION:
JESSICA: I prefer one person be designated IF we can find one person who is willing!
DUSTIN. I believe someone should be appointed as a New Member Director to handle all new member issues.
SUSAN. Here are a few thoughts on that, as well as Dustin's suggestion that we have a New Member coordinator. We do need someone to deal with new members. I volunteer to do the part where I get information from them. As clerk, I need to maintain the list of members and member information, anyway. As regards giving them information about us, I can prepare something to e-mail out regarding our charter, bylaws and incorporation status. HOWEVER -- I would prefer that somebody else answer technical questions. My most common problem with new members is that they e-mail me a list of projects they've worked on or specific items they are interested in pursuing and, quite frankly, I don't understand what they're saying sufficiently to be of any help.
MARK. Once we have a set of goals defined, perhaps we can post what needs to be done and let folks voulnteer for things that need to get done...
IVAN. It depends (I suppose that means I abstain). As has been pointed out by others, the answer to point 4 depends on the answer to point 3: we need to decide what we want to do before we can suggest how people can help. We'll also need to decide how these goals are organised among the workforce. In the meantime, perhaps new subscribers could introduce themselves to the list, along with a paragraph or two on their interests, and then get involved in settling point 3.
It's appropriate that Susan - as Clerk - deal with the organisational side of new subscribers; as for the technical questions that they bring - couldn't they be dealt with by the group as a whole (either from the subscriber in person, or forwarded by Susan)?
Presumably, when ossri.org is up & running there'll be some information on the directors, and on the (kinds of) projects we're interested in.
ED. Motion that we approve Dustin's suggestion with Susan as new member coordinator!
Regarding technical questions, I nominate Christophe to fill this role. If he needs to be a director to do so, perhaps we have a slot available? It would be good to see him take on a prominent role in the organization; his outstanding technical background and qualifications are simply too good not to make full use of. I'd encourage every director to read his resume. I'm not into flattery, but we couldn't ask for a member with stronger technical qualifications to put into a leadership role and enhance the
credibility of this organization. I would be willing to step aside and let him take my place as director if he's interested, if it were necessary, and if approved by the other directors or membership, as needed.
SUSAN--NOTE. Comment to Ed. We can always increase the number of Directors provided that we modify the by-laws at the next meeting. If we feel we need an odd number of voting directors, let's make the disabled representative, Tris Metcalf, voting also.
5. LEGAL
Goodwin Procter has not gotten back to Susan as of today. However, there is no reason to believe that they will not agree to represent OSSRI. All legal actions (incorporating in Massachusetts USA, applying for tax-exempt status at the US federal and state level) are therefore on hold. We expect that within the next week or two we will receive an engagement letter, which Susan has been authorized to sign on behalf of OSSRI.
DISCUSSION:
JESSICA: Thanks, Susan, for all this work.
SUSAN. Goodwin Procter is still very anxious to have us as a pro bono client.
IVAN. I second Jessica's remarks here.
ED. No rush, sounds like waiting is OK.
OTHER ITEMS.
MARK. Do we have an IRC or an IM channel we can use for live discussions? also for collaborative communications a wikki would would be useful (I've had good success with tiki and sharepoint) If we don't already have something setup somewhere I'd be happy to setup tiki on a box somewhere.. finally for meetings, product demonstrations etc.. it's hard to find a better tool than webex.
SUSAN. Reminder. Please follow http://www.ieee-isto.org/ thread.
Meeting Adjourned 27 February 2004 at 13:06 GMT.
Susan R. Cragin
Clerk
Corrections welcome.
("OSSRI")
Thursday, 26 February 2004
Location: OSSRI listserve
and 312 Washington St 4-1, Wellesley MA 02481 USA
The meeting came to order late, at 07:49:12 -0500, or 2:49 GMT.
ATTENDING
DIRECTORS:
Susan Cragin
John Dowding
Jessica Hekman
Volker Kuhlmann
Tristram Metcalfe
Edwin Suominen
Ivan Uemlianin
Dustin Wish
MEMBERS:
Mark Kyes
Turner Rentz III
GUESTS:
None.
OPENING REMARKS:
SUSAN. Our president, Dustin, is incapacitated and I have taken over running the meeting.
DUSTIN. Thank you for calling the meeting to order. I have over 101 degree fever and have been in bed since yesterday morning. I just got up to check my email.
IVAN. First of all, thanks to Susan for taking over the meeting at such short notice!
ED. Note to Dustin: sorry I didn't get you the info. It can wait for the next meeting. Have been fighting recurrent pneumonia. I think you, of all people, will understand right now!
AGENDA:
1. NOMINATION OF MEMBERS.
1.) Saravana Krishnamurthy, Manager-Engineering, SS8 Networks, Inc. from San Jose, CA, USA
2.) Christophe Gerard, NMS Communications, Shrewsbury, MA, USA
3.) Alfred Dayton, Bay Shore, NY
4.) Turner Rentz III
5.) Dennis A. Beckley
6.) Bruce Cyr
VOTE: CONSENSUS
DISCUSSION:
IVAN. Welcome, new members! It might be a nice idea for newcomers to introduce themselves to the list (see point 4 below). That is, unless we want to feel like a big mutlinational corporation.
TURNER. This looks like a well rounded council. I am focussing on Accent recognition.
ED. We are fortunate to have their interest and good will.
VOLKER. Do we really need to vote on each new member, isn't the OSSRI open to anyone?
SUSAN -- NOTE. Perhaps in future, new members should be a non-voting issue.
2. PURCHASE AND HOSTING FOR OSSRI.org
I (Dustin) have personally volunteered to provide hosting for the site. We have a member that volunteered to purchase the domain name for OSSRI. I would like to see the membership discuss this before the meeting so we may get this issue to vote during the meeting.
VOTE: YES-8; NO-0
DISCUSSION:
MARK. Both of these offerings sound great. for the domain, I think the Adminstrative, technical and billing contacts can be assigned to roles that we can define people to fill those roles as needed later..
TURNER. The word "opensource" will soon become one word in the parlance, and the dns osri.org is easier to remember and type. "ossri" sounds like you're saying "sorri" >:)
IVAN. YES to ossri.org (not in favour of the osri.org suggestion)
3. PLAN OF ACTION.
We should have a preliminary plan of action for the technical side. Discussion? Thoughts?
JESSICA. Though I haven't done any work on them lately, I am affiliated with the xvoice and xvoice-sphinx projects. I would love to have people take a look at xvoice-sphinx.sourceforge.net and see if that's a good base for us to start with. xvoice-sphinx was researching getting a pre-existing open source engine to reliably recognize dictated speech. As of the time we gave up, we were unable to get Sphinx to perform reliably. It might be worth a) investigating Sphinx more -- more details on that web page -- or b) looking into other engines.
DUSTIN. Once the site is setup we can link to all major projects and start a CVS code area for the start of our own. The next step is the programming....
MARK. Definately, I think it prudent to come up with a clearly defined set of goals, once the goals have been defined, we should form teams that are responsible for attaining those goals, these teams wouold evaluate technologies, define architecture and craft designs for the goal(s) they've been assigned (just enough so that the larger tasks can be broken off if need be and distributed to others to work on). It's my belief that our process should be as nimble and as light as possible so that we don't get bogged down in to much process but can keep the project moving and our codebase growing. Finally, we'll need a person (maybe more).. that looks at the sytem from a higher level this person/group ensures that the architectures being proposed by the teams are compatible with each other. he/she/they spread good ideas from group to group and becomes the resource the teams go to when they disagree. Finally, I think the teams should strive to come up with a very high level plan of what capability they are looking to have and when it will be available, if for no other reason than they have something to track to and other groups will know when capability will be available for them to use..
IVAN. As Jessica and Mark have suggested, I think a good early goal would be to evaluate the technologies currently available, upon which an open source ASR system could be based. This would mean doing some kind of comparative assessment of Sphinx, ISIP, JULIUS, and no doubt there are others.
One difficulty in carrying out such an assessment might be the lack of training data and/or acoustic models, so we might need to trawl around to see what training data is available at a reasonable price (e.g. free). The above would teherefore inculde evaluation of acoustic model and language model building tools.
It might be worth considering what is the minimal useful functionality for an ASR application: we might come up with something achievable, even if it's only a demo. I'm thinking along the lines of a plug-in for Mozilla or OpenOffice.org - would it be useful to have something that could recognise the names in your address book? or something that would allow a user to navigate GUI menus by speech? (This might mean talking to some end-users;)
Finally, I am no legal authority (slight understatement), but I have a hunch that 'interesting' IP issues may crop up - beyond the 'GPL or BSD?'. For example, some material (e.g., training data from ELDA; various corpora and/or pronunciation dictionaries) is available either only for, or more cheaply for, non-commercial use: if we used this material to build language or acoustic models would that constrain who could use (or whether we could sell) the models (or application using them)? This question may just demonstrate my ignorance of the law, of course, but it does seem 'these days' that IP law is a fairly important component in open source projects. At the very least I'm interested in earning more about this area.
So, some ideas:
(a) evaluate available ASR software
(b) source available training data (for acoustic and language models)
(c) build a library of minimal use cases
(d) build a library of IP issues and their implications
TURNER. The group should focus, in my view, on the technical aspect by focusing on three things. specification, specification, and the specification.
A good plan going forward from that, is to divide up the group into XP units and deal out specific parts based on the dilbert principle.
4. SUBSCRIBERS
We now have 40 subscribers to the listserv. Several contacted Susan off-list to ask how they can help. Should she refer them to all the directors, or would they prefer that one person be designated?
DISCUSSION:
JESSICA: I prefer one person be designated IF we can find one person who is willing!
DUSTIN. I believe someone should be appointed as a New Member Director to handle all new member issues.
SUSAN. Here are a few thoughts on that, as well as Dustin's suggestion that we have a New Member coordinator. We do need someone to deal with new members. I volunteer to do the part where I get information from them. As clerk, I need to maintain the list of members and member information, anyway. As regards giving them information about us, I can prepare something to e-mail out regarding our charter, bylaws and incorporation status. HOWEVER -- I would prefer that somebody else answer technical questions. My most common problem with new members is that they e-mail me a list of projects they've worked on or specific items they are interested in pursuing and, quite frankly, I don't understand what they're saying sufficiently to be of any help.
MARK. Once we have a set of goals defined, perhaps we can post what needs to be done and let folks voulnteer for things that need to get done...
IVAN. It depends (I suppose that means I abstain). As has been pointed out by others, the answer to point 4 depends on the answer to point 3: we need to decide what we want to do before we can suggest how people can help. We'll also need to decide how these goals are organised among the workforce. In the meantime, perhaps new subscribers could introduce themselves to the list, along with a paragraph or two on their interests, and then get involved in settling point 3.
It's appropriate that Susan - as Clerk - deal with the organisational side of new subscribers; as for the technical questions that they bring - couldn't they be dealt with by the group as a whole (either from the subscriber in person, or forwarded by Susan)?
Presumably, when ossri.org is up & running there'll be some information on the directors, and on the (kinds of) projects we're interested in.
ED. Motion that we approve Dustin's suggestion with Susan as new member coordinator!
Regarding technical questions, I nominate Christophe to fill this role. If he needs to be a director to do so, perhaps we have a slot available? It would be good to see him take on a prominent role in the organization; his outstanding technical background and qualifications are simply too good not to make full use of. I'd encourage every director to read his resume. I'm not into flattery, but we couldn't ask for a member with stronger technical qualifications to put into a leadership role and enhance the
credibility of this organization. I would be willing to step aside and let him take my place as director if he's interested, if it were necessary, and if approved by the other directors or membership, as needed.
SUSAN--NOTE. Comment to Ed. We can always increase the number of Directors provided that we modify the by-laws at the next meeting. If we feel we need an odd number of voting directors, let's make the disabled representative, Tris Metcalf, voting also.
5. LEGAL
Goodwin Procter has not gotten back to Susan as of today. However, there is no reason to believe that they will not agree to represent OSSRI. All legal actions (incorporating in Massachusetts USA, applying for tax-exempt status at the US federal and state level) are therefore on hold. We expect that within the next week or two we will receive an engagement letter, which Susan has been authorized to sign on behalf of OSSRI.
DISCUSSION:
JESSICA: Thanks, Susan, for all this work.
SUSAN. Goodwin Procter is still very anxious to have us as a pro bono client.
IVAN. I second Jessica's remarks here.
ED. No rush, sounds like waiting is OK.
OTHER ITEMS.
MARK. Do we have an IRC or an IM channel we can use for live discussions? also for collaborative communications a wikki would would be useful (I've had good success with tiki and sharepoint) If we don't already have something setup somewhere I'd be happy to setup tiki on a box somewhere.. finally for meetings, product demonstrations etc.. it's hard to find a better tool than webex.
SUSAN. Reminder. Please follow http://www.ieee-isto.org/ thread.
Meeting Adjourned 27 February 2004 at 13:06 GMT.
Susan R. Cragin
Clerk
Corrections welcome.