Natural Language Model Re-usability for Scaling to Different Domains
نویسندگان
چکیده
Natural language understanding is the core of the human computer interactions. However, building new domains and tasks that need a separate set of models is a bottleneck for scaling to a large number of domains and experiences. In this paper, we propose a practical technique that addresses this issue in a web-scale language understanding system: Microsoft’s personal digital assistant Cortana. The proposed technique uses a constrained decoding method with a universal slot tagging model sharing the same schema as the collection of slot taggers built for each domain. The proposed approach allows reusing of slots across different domains and tasks while achieving virtually the same performance as those slot taggers trained per domain fashion.
منابع مشابه
Re-usability of traffic signs for inactive drivers with consideration of personal characteristics and sign features
There has been an increasing concern about inactive drivers who would easily lead to road accidents and fatalities once return to driving. This study investigated the re-usability of traffic signs for inactive drivers with consideration of driver factors and cognitive sign features. Fifty-seven Hong Kong Chinese, who possessed a full driving license but had not driven for an extended period, co...
متن کاملLinguistic Knowledge Generator
The difficulties in current NLP applications are seldom due to the lack of appropriate frameworks for encoding our linguistic or extra-linguistic knowledge, hut rather to the fact that we do not know in advance what actual znstances of knowledge should be, even though we know in advance what types of knowledge are required. It normally takes a long time and requires painful trial and error proc...
متن کاملInvited Talk: Domain-adaptation of Natural Language Processing Tools for RE
Natural language processing tools like part-of-speech taggers and parsers are being used in a variety of applications involving natural language, including RE. Such tools, based on statistical models of language, are learnt via supervised machine learning algorithms from human-annotated data. Due to their dependence on annotated data, which is limited in size and genre, these models have a fall...
متن کاملImplicational Scaling of Reading Comprehension Construct: Is it Deterministic or Probabilistic?
In English as a Second Language Teaching and Testing situations, it is common to infer about learners’ reading ability based on his or her total score on a reading test. This assumes the unidimensional and reproducible nature of reading items. However, few researches have been conducted to probe the issue through psychometric analyses. In the present study, the IELTS exemplar module C (1994) wa...
متن کاملReversibility and Re-usability of Resources in NLG and Natural Language Dialog Systems
Reversibility is a key to efficient and maintainable NLG systems. In this paper, we present a formal definition of reversible NLG systems and develop a classification of existing natural language dialog systems in this framework.
متن کامل