Google Search Quality: Global and Local
- From RSS
Daphne Dembo, Development Director
The goal of Google is to make the search service effective for each user, no matter where he lives and what language he speaks. Many of the algorithms that we usually use can be applied to all languages, but in some cases unique characteristics require a special approach.
Today, Google searches in more than a hundred languages. At first, engineers of the department for improving the quality of processing search queries were engaged in the development and improvement of the search system on each of them. Each specialist enthusiastically worked on solving specific problems for his language. As the volume of foreign language searches grew, we turned for help to company employees around the world. Now a large team of specialists is working on the problems of improving the search, which help us understand the unique features of languages.
To overlay a search model in a particular language, the first thing we do is test the search on a variety of documents in that language. In this way, we can make sure that the language model is accurate and universal, for example, that it includes names, idioms, vernaculars, and neologisms that are not yet included in dictionaries. After the general setup, indexing the documents, we get the opportunity to categorize and parse them in accordance with the user's request. After this step is completed, the localized search is launched.
In addition to language features, to provide the most relevant results, the search system must take into account the user's location, at least with the accuracy of the country. A request from [Côte d'Or] in France should provide information about this region, and in Belgium about a chocolate producer. The information must also comply with the standards of the country. For example, at the request of the Norwegian user [ Været i Oslo ], the temperature will be shown in degrees Celsius, and for the American [ weather Boston ], in degrees Fahrenheit.
What Google can do
Today, Google Search is being developed for more than one hundred languages in one hundred and fifty countries of the world, and many improvements are launched every month. The above options are just a small part of what we work on every day to make the search even better. Ease of use, the appearance of the start page and the search results page, understanding the cultural characteristics of a particular country is also extremely important for creating a search engine that will be close and understandable to people from different countries.
The goal of Google is to make the search service effective for each user, no matter where he lives and what language he speaks. Many of the algorithms that we usually use can be applied to all languages, but in some cases unique characteristics require a special approach.
Today, Google searches in more than a hundred languages. At first, engineers of the department for improving the quality of processing search queries were engaged in the development and improvement of the search system on each of them. Each specialist enthusiastically worked on solving specific problems for his language. As the volume of foreign language searches grew, we turned for help to company employees around the world. Now a large team of specialists is working on the problems of improving the search, which help us understand the unique features of languages.
To overlay a search model in a particular language, the first thing we do is test the search on a variety of documents in that language. In this way, we can make sure that the language model is accurate and universal, for example, that it includes names, idioms, vernaculars, and neologisms that are not yet included in dictionaries. After the general setup, indexing the documents, we get the opportunity to categorize and parse them in accordance with the user's request. After this step is completed, the localized search is launched.
In addition to language features, to provide the most relevant results, the search system must take into account the user's location, at least with the accuracy of the country. A request from [Côte d'Or] in France should provide information about this region, and in Belgium about a chocolate producer. The information must also comply with the standards of the country. For example, at the request of the Norwegian user [ Været i Oslo ], the temperature will be shown in degrees Celsius, and for the American [ weather Boston ], in degrees Fahrenheit.
What Google can do
- If the user forgets to switch the input language and the request becomes meaningless, Google will correct the error. For example, the Cyrillic query [ rehc ljkffhf ] will be corrected to [ dollar exchange rate ].
- A Google search “understands” transliterated words. For example, on request in Latin [ biskvitnyi rulet ], the user receives the results in transliterated [ biscuit roll ].
- Google search is trying to save you time. The Google Suggest service offers you query options by typing a word online.
- Google can translate the request into English, then search the web for documents and translate the results into the original language of the request. For example, Korean users can search for Harry Potter news for [ 해리 포터 ], while Arabic users can search for Sydney Opera House news for [ افتتاح دار الاوبرا في سيدني ].
- Google will fix grammar errors. For example, if you don’t remember how to accurately write “smoke detector” in Estonian, the system will offer a corrected version for the word [ suitsuantur ].
- Google will take into account diacritics: the search engine will yield equally good results if the query is entered with diacritics [ Météo à Québec ] or without them [ meteo quebec ].
- Google can recognize synonyms in different languages. Samsung is a synonym for Korean "삼성", so users who request [ samsung ] get results containing the company name in Korean.
- Recognizes complex words: documents dedicated to the University of Uppsala in Sweden will be issued upon request by the word [ uppsalauniversitet ], and by a separate phrase [ Uppsala universitet ].
- Google searches for derivatives from one root. Polish user can search for the answer to a query for the word "film» [ film ], and get results that contain the root of the word, in particular - «filmów», «filmu» , «filmie», «filmy».
Today, Google Search is being developed for more than one hundred languages in one hundred and fifty countries of the world, and many improvements are launched every month. The above options are just a small part of what we work on every day to make the search even better. Ease of use, the appearance of the start page and the search results page, understanding the cultural characteristics of a particular country is also extremely important for creating a search engine that will be close and understandable to people from different countries.
