Also featuring in the list of inventors are Alex Franz
and Brian Milch who co-authored a paper in 2002, Searching the
Web by Voice, which considered a speech interface for Google
for settings where typing on a keyboard is impractical.
The patent application reflects the paper. The difficulty of
applying speech recognition to web searches is explained: most
queries are short, just five or six words; and a large vocabulary
is needed – yet even a vocabulary of 100,000 words only covers
about 80% of the query traffic. The paper addresses the problems
and solutions such as a likelihood ratio, applying this to
collocations (e.g. a search on "New York flowers" should combine
New York + flowers, not New + York Flowers), constructing language
models, analysing perplexity results, etc.
The paper suggested that a commercial speech recognition engine
could return the correct transcription of a spoken search query
among its top 10 hypotheses about 60% of the time.
According to the abstract of the patent, entitled Voice
interface for a search engine:
"A system provides search results from a
voice search query. The system receives a voice search query from a
user, derives one or more recognition hypotheses, each being
associated with a weight, from the voice search query, and
constructs a weighted boolean query using the recognition
hypotheses. The system then provides the weighted boolean query to
a search system and provides the results of the search system to a
user."
