Dr. Hossein Eslambolchi
Information overload is nothing new. As long as we have been recording information, we have been inventing ways to retrieve it.
The Hittites added title page details to documents back in the thirteenth century BCE. Pliny the Elder in the year 77 included a table of contents and a bibliography in his landmark work The Natural History. Even ancient libraries classified their scrolls and tablets. By the year 1500 with more than nine million printed books in circulation – about the size in terabytes of the print collection of today’s Library of Congress – medieval librarians developed new classification systems to organize knowledge.
Today the amount of information available to us on the World Wide Web is truly extraordinary. You can find everything from the full text of Cicero’s speeches to the Roman Senate, the lyrics to a Cole Porter tune, a photograph of the Grand Canal, a repair manual for a 1967 Chevy, or a lecture series on the history of jazz music complete with audio files and video.
Search engines, portals, and online directories constantly improve in order to help us access this flood of information: improved search algorithms, new techniques for locating non-textual resources, and the bundling of related applications such as online shopping and communications services. But as the information on the World Wide Web continues to expand in size and diversity, it is ever more challenging to find what you are looking for.
Formatting an efficient search query still requires a lot of experience — or a willingness to experiment. When we search, we typically don’t spend very much time seeking an answer; in fact, we often don’t look beyond the first page of search results. Few of us bother to learn the command structures and more advanced features that will help refine a query, or try out some of the newer services that creatively organize search results.
Creative innovations are helping to refine search results, both simplifying and hastening the search process. Some search interfaces allow you to dynamically modify the parameters of your search by moving “sliders” up or down a scale to add or reduce the importance of certain ranking factors thereby refining a search. Visual presentation of search results offers an alternative way to see relationships within the various sub-topics. Tagging can speed the location of new information.
But by 2020, we will most likely look back at these innovations as simply incremental improvements compared with what will be possible then.
Using Natural Language
We’re all familiar with the “ifs”, “ands,” and “ors” of Boolean searching and the use of keywords to find what we want. The more adventurous of us use a search engine’s special features to set the parameters of what is wanted, such as data in an excel spreadsheet or information from a recently updated web site. Most search engines use natural language processing to process queries, so we can also try entering a query in the form of a question. But the quality of results from entering a query in natural language can vary dramatically by the type of question and the search service used.
In the future searchers will routinely enter their queries in the form of a question or a statement just as they would speak to another human being. The search technology will be able to parse the query to determine what is needed and will be able to deal with quite complex questions that may require accessing a wide variety of different sources with multiple media types. Instead of entering a list of keywords, searchers will speak conversationally, with questions such as “Is there a Greek restaurant nearby” or “What killed the dinosaurs” or even “Find me a recording of Beethoven’s Trio in B Flat by the group that played at Carnegie Hall last night.”
Future search systems are likely to include an interactive component using natural language questioning which will query the searcher in order to limit the universe of possible data sources and provide greater precision in final results. When asked a question such as “What happens when a comet hits the earth?, the system may ask in turn “Any comet or one large enough to cause major destruction?” A complex question such as “What are the roots of modern economic theory that can be traced to the Scottish Enlightenment?” could result in the reply “Would you like to focus on the theories of Adam Smith?”
The technology underlying the search interface will include sophisticated reasoning systems that can make inferences from the searcher’s situation and from the system’s collective experience. The search system will be able to build on its accumulated knowledge just as human beings make judgments based on their own lifetime experience and observation. It will respond to new questions based on previous search requests as well as deriving meaning from associated experiences. This automated reasoning process will include the ability to differentiate commonly accepted fact from opinion and to reason out the appropriate level of detail needed to sufficiently answer the searcher’s query.
Search systems will be able to judge from the form and subject of the question, as well as other interactions with the questioner, what level of detail is necessary — and the best way to format the answer. A query such as “list the key symptoms of Type B diabetes” is likely to require a more sophisticated answer than the query “Who are the Seven Sisters” sent from a cell phone. Search systems will be able to make judgments as to the questioner’s intent. From a query such as “what are the safest mid-sized cars?” the system would conclude that the searcher is likely to be in the market for a new car and include a list ranked by price of cars for sale near her home.
Future search systems will also be far more sophisticated in how they present the results of a query. Rather than producing a ranked list of sources, the search engine will detail an answer to a simple question or summarize the results in a report organized by clusters of information and sources on different aspects of the topic. At this point the searcher will be able to ask additional questions, purchase an item, make a call to a listed expert, share the results with a group, or merge the information with another document.
Future generations of search technology will also add a richer contextual component to the searching process. In addition to helping us describe what we want to find, they will help us identify our level of interest and expertise, derive how information should best be presented, and infer how we will access it.
Eventually search systems will make use of sensors that will automatically associate information points to present answers in the context of our past activities, interests, and other aspects of our lives. A query regarding an audio recording, might result in the system reminding us that we might already own a copy as it is listed on the CD inventory we maintain in our entertainment hub at home. The search system will be able to correlate a question about diving equipment to past purchases of similar items as a birthday gift for a relative, automatically providing the date when the item should be shipped in order to arrive before he blows out the candles.
We are seeing early examples of information presented in context with some of today’s mobile searching capabilities. Cell phone users, for example, access web pages specifically designed for the screen of a handheld device. Most of the general purpose search services have local search services where users can find information related to a specific city or zip code. Users can search for restaurants, stores, and other businesses in a specific area, obtain driving directions and maps, and even sometimes use a click-to dial feature to make a telephone call. Users can even bypass web searching altogether by sending and receiving SMS queries.
In the future, search services will automatically be able to discern the best way results should be delivered. They will know whether we are on the phone, at home talking to a friend, or working on a project in the office — they will draw conclusions from the limited information that we, or the ambient circumstances, provide. They will be able to postulate scenarios of possible meaning and intent and use commonsense reasoning to automatically present results. They will be able to discern information about a questioner’s background from her accent or use of variant spelling and combine this knowledge with her current location to present information in context. Shopping information might be presented in dollars and euros if a Scottish man is asking a question about raincoats from a location in Maine. The response to a question about what is playing currently London theaters would be predicated on other interactions indicating that the inquirer is on vacation in England and is fond of Jacobean literature.
By 2020, search systems will be able to make quite sophisticated contextual inferences. By then search technologies will have developed far beyond the ability to take into account geographic location or personal preferences, but will make inferences based on perceived emotion, cultural perspective, level of topical knowledge, and more. A query about local hospitals might result in driving directions and a click-to-dial 911 button in response to the tension perceived in the questioner’s voice. A question about flight schedules will be able to discern from ambient noise that we are in an airport and will adjust the data presentation accordingly. Future search systems will make correlations among a wide range of data points and our accumulated electronic history. Recipes will be translated into grams for a transplanted European making a cake in Canada. From past experience, a search system will be able to judge that an inquirer can read classical Latin but will need translations for sources in ecclesiastical Latin. An inquiry about batting averages sent from Yankee Stadium in the middle of the World Series will receive a quite different answer than a search performed by a fourth-grader working on her math homework.
By the year 2020, search as we know it will be an artifact. We won’t connect to a web site to search for information. Instead, services will incorporate search capabilities. Information agents will be embedded into these services and will answer questions, make suggestions, update us on past queries and provide us with reminders. They will anticipate our needs based on our personal preferences and the tasks we are engaged in.
When we do conduct a search for information, instead of entering keyword queries into a search interface and sifting through a list of static links, we will interactively converse with information agents that will proactively compile data from multiple sources. The search system will use its reasoning capabilities to comprehend what we are requesting and how to deliver the answer. Boolean logic will be once again relegated to algebra class.