Human Language Technologies

The Human Language Technologies (HLT) department at IBM Research - India aims to advance the science of processing, analyzing and understanding human languages expressed in various modalities and applying it to create differentiated products and services for IBM and benefit the society at large.

We deal with unstructured, noisy, and multi-lingual information sources such as Web pages, documents, e-mails, ticketing systems, team rooms, call center calls, etc. The areas of research include natural language processing, information retrieval and extraction, text analytics, data mining, machine translation, and speech processing.

We build analytical models and solutions to generate business intelligence and insights that can improve the quality, productivity and effectiveness of IT and IT-Enabled services. We partner with IBM’s internal business units – these include Contact Center Services, Application Development Services and Infrastructure Services. We also work directly with IBMs clients in a variety of industries such as banking and retail, among others.

One of our focus areas is Text Analytics. We are working on various applications of deep text analytics, natural language processing, information extraction, information retrieval, machine translation, and machine learning for services businesses. Examples include developing applications of focused language translation for services, improving the organization and reuse of knowledge extracted from previous interactions to help application and IT support teams, automatically evaluating and filtering resumes of people applying for jobs using IR/IE techniques, etc.

In the area of speech technologies, we are working on acoustic modeling, language modeling, embedded grammars, pronunciation & fluency evaluation and audio mining. In the area of Contact Center research, we are developing an array of technologies ranging from automatic evaluation of the linguistic abilities of applicants to customer satisfaction analysis. Contact center analytics uses heterogeneous data sources such as audio conversations, agent logs, and customer satisfaction surveys in order to increase productivity and generate greater insights into the products and services.