Language, Speech and Multimedia Technologies Observatory
10/22/2010 - 07:55

Photo courtesy: Flickr/jurvetson

It’s likely you’ve heard it often enough from friends and family trying to find a job in a country whose unemployment rate, as of Oct. 8, refused to budge lower than 9.6 percent: “I’m the perfect candidate for these jobs I’ve applied to. But it’s like my résumé goes into a black hole, and I never hear from the recruiter.”

Well, their résumé might not be falling into a black hole, but it certainly is a deep one. It’s bad for the job-seeker but bad for the potential employers, too, who may very well miss out on hiring the talent that would do the job best. That’s why we’ve seen job search monsters (literally) like start leveraging semantic technology to make HR’s job easier to do (see here ).

In fact, Monster Worldwide Inc. late last month wracked up its third patent or its 6Sense Semantic Search technology, which supports its Monster Power Resume Search and Candidate Spotlight, as well as Seeker Job Search. Said Darko Dejanovic, Global CIO and head of product for Monster Worldwide, in a statement announcing the news, “This multi-patented recognition reinforces that our semantic search will continue to be the cornerstone in a broader strategy to bring increased efficiency, transform the recruiting process and grow our business.”

But Monster’s not the only one hoping to realize such returns from helping companies accelerate recruitment business processes. While it’s unclear if Bintro, an early entry in this field, is any longer active, this week brings news that TalentSpring, a provider of semantics-based candidate sourcing services, was acquired by Talent Technology Corporation. Though we’ve recently discussed the tightening VC market (see this week’s story), TalentSpring proved its merit to the tune of $1.6 million last April when it received an investment from Second Avenue Partners and private investors to take to the market its service for automatically searching social networking sites, job boards, and corporate applicant tracking systems to match job candidates with open positions.


New Career Opportunities Daily: The best jobs in media.
10/18/2010 - 14:35
Gaia: Contextual salience in query-based summarization
Tokia: 2.2 mintegia
Hizlaria: Wauter Bosma (Vrieje Universiteit Amsterdam)
Eguna: Urriaren 19a Ordua: 15:00

Amsterdameko Unibertsitate Libretik
etorri zaigu bisitan Wauter Bosma ikerlaria. Hilebetez ibiliko da
gurekin lanean, gu ere partaide garen Kyoto proiektuan lankidea dugu Wauter.
Bere ikerlerro nagusia laburpena automatikoa da. Eta horretaz hitz egingo du bihar asteartean. Berak teknika berri bat garatu du laburpenak grafuen bidez automatikoki lortzeko, beti ere testuinguruan azaltzen diren kontzeptuen arteko harremanak asimilatzeko asmoz.

Discourse theories claim that text gets meaning in context. Most
summarization systems do not take advantage of this. They assess the
relevance of each passage individually rather than modeling the way
context affects the relevance of passages. In order to model relations
in text, I developed a framework for graph-based summarization, so that
the passages can be viewed in a broader context. The result is a
summarization system which is more in line with discourse theory but
still fully automatic. I evaluated the content selection performance of
an implementation of the framework in different configurations.  The
system significantly outperforms a competitive baseline (and participant
systems) on the DUC 2005 evaluation set.
10/18/2010 - 08:50


Courtesy: Flickr/Sahaja Meditation

Customer service takes place across many channels these days – the traditional call-in contact center, but also over the Internet through chat, email, and social media. Companies are adapting to that new reality, but perhaps not yet leveraging as well as they might the insight they’re gaining from these many mediums in ways that can benefit all customers’ experiences, rather than just solve problems on an individual basis.

Perhaps that can start to change, as text and sentiment analytics vendors pair up with some of the big players with software for analyzing customer interaction recordings – which is starting to happen with last week’s announcement that Clarabridge and call center workforce optimization vendor Verint have paired up their technology capabilities to understand interactions across talk and text channels.


New Career Opportunities Daily: The best jobs in media.
10/15/2010 - 14:15

2010/10/15 -- Jeff Allen
10/15/2010 - 08:00

Posted by Mike Schuster & Kaisuke Nakajima, Google Research

Google Voice Search has been available in various flavors of English since 2008, in Mandarin and Japanese since 2009, in French, Italian, German and Spanish since June 2010 (see also in this blog post), and shortly after that in Taiwanese. On June 16th 2010, we took the next step by launching our Korean Voice Search system.

Korean Voice Search, by focusing on finding the correct web page for a spoken query, has been quite successful since launch. We have improved the acoustic models several times which resulted in significantly higher accuracy and reduced latency, and we are committed to improving it even more over time.

While voice search significantly simplifies input for search, especially for longer queries, there are numerous applications on any smartphone that could also benefit from general voice input, such as dictating an email or an SMS. Our experience with US English has taught us that voice input is as important as voice search, as the time savings from speaking rather than typing a message are substantial. Korean is the first non-English language where we are launching general voice input. This launch extends voice input to emails, SMS messages, and more on Korean Android phones. Now every text field on the phone will accept Korean speech input.

Creating a general voice input service had different requirements and technical challenges compared to voice search. While voice search was optimized to give the user the correct web page, voice input was optimized to minimize (Hangul) character error rate. Voice inputs are usually longer than searches (short full sentences or parts of sentences), and the system had to be trained differently for this type of data. The current system’s language model was trained on millions of Korean sentences that are similar to those we expect to be spoken. In addition to the queries we used for training voice search, we also used parts of web pages, selected blogs, news articles and more. Because the system expects spoken data similar to what it was trained on, it will generally work well on normal spoken sentences, but may yet have difficulty on random or rare word sequences -- we will work to keep improving on those.

Korean voice input is part of Google’s long-term goal to make speech input an acceptable and useful form of input on any mobile device. As with voice search, our cloud computing infrastructure will help us to improve quality quickly, as we work to better support all noise conditions, all Korean dialects, and all Korean users.;5715a7f.1010b
10/14/2010 - 07:00

Dear all,

[apologies for cross-postings]

I've compiled a list of conferences/deadlines with relevance to our community. I aim to send an updated overview every once in a while.

There's a printerfriendly version for hanging on your wall at

First, here's a list of popular conferences that are *missing*.
If you know someone from the organizing team, please point them to our calendar [...]
10/14/2010 - 07:00


"Bringing MT to the User: Research on Integrating MT in the Translation Industry"
Second Joint EM+/CNGL Workshop (JEC 2010)

At the AMTA 2010 conference (, the EuroMatrix+ Project ( and the Centre for Next Generation Localisation ( are organising the Second Joint EM+/CNGL Workshop, titled "Bringing MT to the User: Research on Integrating MT in the Translation Industry". The workshop will take place in Denver, Colorado on 4 November 2010, immediately after the main AMTA 2010 conference.

Recent years have seen a revolution in MT triggered by the emergence of statistical approaches to MT and improvements in translation quality. MT (rule-based, statistical
10/08/2010 - 07:20

First International Conference on Robot, Vision and Signal Processing [Kaohsiung City, Taiwan] [Nov 21, 2011 - Nov 23, 2011]
10/08/2010 - 07:20

Agreement focused on advancing natural language processing technologies to drive evidence-based care and improve patient outcomes.
10/08/2010 - 07:20

2010/10/07 -- Alon Lavie

Syndicate content