Obviousness

Obviousness Analysis of U.S. Patent 9,269,097

As a senior patent analyst, this analysis evaluates the obviousness of U.S. Patent 9,269,097 ('097 patent) under 35 U.S.C. § 103. This statute precludes issuing a patent if the differences between the claimed invention and the prior art are such that the invention as a whole would have been obvious to a person having ordinary skill in the art (PHOSITA) at the time the invention was made. The priority date of the '097 patent family is February 6, 2007, which establishes the critical date for evaluating the state of the art.

The analysis hinges on whether a PHOSITA, with knowledge of the technological landscape in and around 2007, would have found it obvious to combine existing elements of voice recognition, targeted advertising, and contextual analysis to arrive at the invention claimed in the '097 patent. The core of the invention is using a presented advertisement as the immediate context for interpreting a subsequent natural language utterance, particularly for resolving pronouns.

Based on the prior art, a strong case for obviousness can be made by combining references that teach voice-based search with references that teach contextual understanding in human-computer interaction.

Proposed Obviousness Combination 1: Bennett (US 7,058,567) in view of Gildea (US 2005/0288929 A1)

• U.S. Patent 7,058,567 (Bennett): Titled "Voice-based interface for the world wide web," Bennett discloses a system that allows a user to navigate the web and interact with web content using voice commands. It teaches receiving a user's speech, recognizing the speech to produce text, identifying a command, and executing that command in the context of a currently displayed webpage. This establishes the foundational concept of a voice command being interpreted relative to on-screen content.

• U.S. Patent Application 2005/0288929 A1 (Gildea): Titled "System and method for context-based speech recognition," Gildea teaches a speech recognition system that uses contextual information to improve accuracy and resolve ambiguity. Gildea explicitly discusses using prior utterances or dialog states as context. The core teaching is that the interpretation of a user's speech is enhanced by understanding the environment and the preceding interaction.

Motivation to Combine:

A PHOSITA in 2007 would have been motivated to combine the teachings of Bennett and Gildea for a clear and predictable purpose: to improve the user experience of the voice-enabled web navigation system described by Bennett.

1. Solving a Known Problem: A well-known problem in natural language processing was ambiguity. Gildea was directed at solving this very problem by using context. A user of Bennett's system interacting with an advertisement on a webpage might say "buy it" or "call them." This creates an ambiguity that Bennett does not explicitly solve. A PHOSITA would look to the art for solutions to ambiguity and find systems like Gildea's that use context.
2. Predictable Improvement: Applying Gildea's contextual processing to Bennett's system represents a predictable improvement. The advertisement displayed on the webpage in Bennett's system is a form of context. A PHOSITA would have recognized that using this immediate, on-screen context (the ad) to interpret a subsequent voice command (per Gildea's method) would make Bennett's system more robust and intuitive. The "teaching-suggestion-motivation" (TSM) test is met because the prior art itself suggests the combination to achieve a predictable result.

Mapping to Independent Claim 1:

• "providing... an advertisement associated with a product or service for presentation to a user;": This is implicitly taught by Bennett, which describes a voice interface for interacting with web content, which commonly includes advertisements. A webpage displaying an ad for a product is a primary example of the environment in which Bennett's invention operates.
• "receiving, at the computer system, a natural language utterance of the user;": This is explicitly taught by both Bennett and Gildea.
• "interpreting... the natural language utterance based on the advertisement...": Bennett teaches interpreting an utterance in the context of a webpage. Gildea teaches the broader principle of using context to interpret an utterance. The combination makes it obvious to use the specific context of an advertisement on the webpage to interpret the utterance.
• "...and, responsive to the existence of a pronoun in the natural language utterance, determining whether the pronoun refers to one or more of the product or service or a provider...": This is the natural and obvious result of combining Bennett and Gildea. If a user of Bennett's system is viewing an ad for "Product X from Company Y" and says "buy it," a PHOSITA applying Gildea's context-based disambiguation would logically conclude that the pronoun "it" refers to Product X. This is a simple application of pronoun resolution, a known technique in natural language processing, using the most salient and immediate context available (the ad).

Proposed Obviousness Combination 2: Franz (US 7,027,987) in view of Gildea (US 2005/0288929 A1)

• U.S. Patent 7,027,987 (Franz): Titled "Voice interface for a search engine," Franz discloses a system that receives a voice search query, derives recognition hypotheses, and provides the query to a search system. This reference establishes the concept of a voice "front-end" for a larger information retrieval system, like a search engine, which is a common platform for serving advertisements.

Motivation to Combine:

Similar to the first combination, a PHOSITA would be motivated to combine Franz and Gildea to enhance the functionality of the voice search interface. Search engines were, by 2007, inextricably linked with advertising. A user's search query often resulted in a page of results that included sponsored links (advertisements). A subsequent voice command, such as "tell me more about the first one," would be ambiguous without context. A PHOSITA would combine Gildea's context-based interpretation with Franz's voice search system to resolve such ambiguities, where the search results page (containing ads) serves as the context.

Mapping to Independent Claim 1:

The mapping follows the same logic as the Bennett and Gildea combination. Franz provides the system for voice interaction with a search engine that presents ads. Gildea provides the method for using the context of those ads to interpret subsequent commands, including resolving pronouns that refer to the advertised products or services.

Conclusion

The independent claims of the '097 patent describe the application of a general principle (context-based interpretation) to a specific environment (voice interaction with advertisements). The prior art from before 2007, including Bennett, Franz, and Gildea, already taught the core components: voice control of computer systems, interaction with on-screen content including ads, and the use of context to resolve ambiguity in speech recognition. A person having ordinary skill in the art would have been motivated to combine these known elements to produce the claimed invention for the predictable result of an improved and more intuitive user interface. Therefore, the claims of U.S. Patent 9,269,097 are vulnerable to a challenge of obviousness under 35 U.S.C. § 103.