LTPC

Goal

The main objective of the workshop is to consider the intersection of two research domains which have been separated the past years: NLP and pervasive computing.
In pervasive and ubiquitous computing scenarios, spoken language is in many cases the ideal modality for human beings to interact with a “disappearing” computer, i.e. to directly formulate their intentions and to receive feedback from the system. However, despite recent advances in speech technology, many developers still have objections to employ NLP due to concerns of low recognition rates and issues of disambiguation etc. These limitations of NLP components are not surprising, considering that even human beings can only make sense of spoken language by contextual knowledge. Sensing context however is one of the key topics of the Pervasive Computing conference series, particularly location sensing and activity recognition. We believe that the pervasive computing community can make an important contribution to the field of NLP and also Human Language Technology (HLT) can profit enormously from pervasive computing by getting contextual information.

Topics of Interest

Some key challenges, on which the workshop will be based on (but are not limited to), include:

•  Location awareness in speech-enabled systems
• Adapting speech systems to the user’s activity
• Multimodal interaction in intelligent environments, e.g. speech and gesture combined
• Interacting through speech with smart devices in the Internet of Things
• Personalizing devices and services through speaker recognition
• Multilingualism, localization, and translation in pervasive computing

The location of users within the environment and the surrounding objects are of major importance for dialogue between the user and a system. It is more likely that the user refers to objects that are nearby and visible, so the speech system should consider this. The challenge is to sense and represent the position of the user and other objects in a location model, and to incorporate this model into the language processing. Likewise, utterances are likely to refer to the current actions of the user. Much effort has been spent in recognizing and representing activity and NLP components should make use of such contextual information.
Gesture recognition is another pervasive technology that can improve speech-enabled systems through multimodality, with manifold applications. Smart energy scenarios drive the connection of objects with the Internet in order to measure and reduce power consumption. This development offers new chances to remotely control and even interact with objects through speech. Likewise, our ageing society requires technological solutions that allow elderly users to live independently at home (Ambient Assisted Living). Such assistance systems require intuitive user interfaces, i.e. based on speech, to keep the humans in the loop.
The quality of user interfaces may be further increased by personalization through speaker recognition.
Finally, user interfaces of pervasive computing systems must be translated and localized to different languages and cultures in order to be successful on global markets. Hence NLP frameworks should include tool support for efficient and correct translation of resources, such as grammars.
We welcome ongoing work on HLT-Pervasive Computing, but also theoretical discussions on how to integrate these areas.