Skip to content 
Google is by far the most important search engine in almost all countries of the world, but is still working on permanently improving its results. Last October, among other things, the Update Bert which helps to better classify the context of individual words in long search queries. Now Google has developed the conversational search which interprets the contextual understanding over several consecutive search queries and thus should help to understand the actual search intention even better.
In one example, Google explains the Conversational Search using a conversation between two people, during which the conversation partners exchange different questions on the same topic. Of course, the people do not forget the context of the questions and the information they have exchanged so far, which leads to further questions being answered in a context-sensitive way. In the future, Google searches should be able to do this.
Search results consider history
According to Google, the Conversational Search provides better results, especially for search queries that allow different interpretations. The search engine makes this clear using the example of Thanksgiving. Users who search for a turkey recipe in the first query will get relevant results because the search query has a clear search intention.
The user then searches for "carving". Google could now either display results that deal with the topic of woodworking (carving, chiseling, cutting up) or Instructions to cut up a turkey. Up to now, the results have mainly been output using the most common interpretation. Conversational Search now ensures that previous search queries are also taken into account and that in the example with "carving" the user is most likely searching for information about carving a turkey.
Relevant information on related topics
In addition, the new language understanding function recognizes search queries within a specific subject area. This is supposed to automatically display further information from relevant subject areas. In one example, Google explains that queries for the films "Polarexpress" and "A Christmas Story" were most likely made because the user is looking for family-friendly films for Christmas. A list of other relevant titles is then automatically displayed in the results.
Contextual learning is also intended to help the user with relevant follow-up questions. For this purpose, related searches for specific questions are already displayed in the "Similar questions" section. In the future, the questions displayed there should refer to the previous searches and not only consider the concrete question.
