Smart Search enhances the Web of Science search by offering an intuitive, intelligent and personalized search experience. As users begin typing, it provides suggestions and comprehends their intent, whether they use natural language or specific search terms. Users can enter search terms without needing to master complex Boolean syntax, as the system constructs the Boolean query in the background.
How Smart Search works
When you enter a query in Smart Search, it goes through a multi-step process to deliver the most relevant results. First, the query is analyzed by our Natural Language Processing (NLP) Parser, which understands the structure and context of your input, identifying key entities such as author names, topics, or research IDs. This helps ensure the search is accurately targeted. The parsed information is then used to generate a structured Web of Science (WOS) query, which is sent to our powerful search engine.
Smart Search performs two types of searches: Document Search and Author Simple Search, handling document and author queries separately. For document searches, if the query is semantic, AI-driven technology is used to interpret the meaning and context of the search, while non-semantic (Boolean) queries follow a standard keyword-based approach. All document search results are then ranked by our AI model to highlight the most relevant ones. Similarly, author searches involving topics are enhanced by an AI-based relevance ranking. The final results are presented in two separate tab: DOCUMENTS and RESEARCHERS, so you can easily explore the most pertinent information from each category.
Smart Search features
- One-Box Search is designed for simplicity, just type naturally. The system intelligently handles typos, language toggling, and even advanced query conversion behind the scenes, all while offering a personalised, multilingual, and dual-tab search experience. In One-Box Search, users cannot manually enter advanced Boolean queries, but the system still converts their user query into Boolean logic in the background. If users type AND, OR, NOT, the system interprets them as English words but still structures a Boolean query in the background for the best results. Even if user type AND, OR, or NOT, the One-Box does not treat them as Boolean operator, it sees them as regular words. However, in the background, the system still converts your input into a Boolean query for precise results.
- Understanding user search query/input (Even If You Use ‘AND’ or ‘OR’)
Examples:
| What You Type | How the System Interprets It | Boolean Query It Actually Runs |
| DNA AND RNA | (Understands "and" as a search term and include it in the search) | TS= (DNA AND “and” AND RNA) |
| AI OR machine learning in medicine | (Understands “or" as a search term and include it in the search) | TS= (AI AND “or” AND machine AND learning AND in AND medicine) |
| cancer NOT chemotherapy | (Understands "not" as a search term and include it in the search) | TS= (cancer AND “not” AND chemotherapy) |
- One-Box Search vs. Advanced Search
Advanced Search is a separate search functionality (UI Tab) that maintains the existing search functionalities for experienced users who prefer precise search or complex Boolean queries.
| Feature | Smart Search | Advanced Search |
| Search functionality | One-box search (new) |
All existing search features: Fielded Search (Basic) Query Builder (Advanced) Citated reference search Structured search
|
| Ease of use | Simple, intuitive, user type in natural language (does not support complex Boolean syntax) | Precise, requires Boolean syntax (AND, OR, NOT) knowledge |
| Typeaheads & Autocomplete | Yes, AI-enabled typeaheads (with personalized search history and suggestions) | Yes, selected fields only – not AI-enabled not supported |
|
Autocorrection (showing results ‘for’ instead ‘of’)
Did you mean? |
Yes, both supported
|
Autocorrection not supported Did you mean? supported
|
| Boolean Query Conversion | Automatic (but user doesn’t see it) | Manual (user customize query) |
| Boolean Search result | Yes – on summary page, document tab | Yes – on summary page, document tab |
| Semantic search result | Yes (understand user intent) - on summary page, document tab | No, semantic search not supported |
| Search personalization | Yes (based on user query) | No |
| Translated Search | Yes (Simplified Chinese only) | No |
- Multilingual & Cross-Database Document Search
All Databases Search: One-Box Search allows user to search across all databases in one go, making search experience seamless and comprehensive. Individual silos cannot be searched separately.
WoS Core Collection (WoSCC) Search: One-Box offers the flexibility to search either across All Databases or exclusively within the WoS Core Collection.
Translated Search: One-Box supports simplified Chinese input! When a user enter a search term in simplified Chinese, a language toggle will appear on search results page, allowing user to switch between English and simplified Chinese seamlessly.
On the search results page, all 50 titles will be automatically translated to Simplified Chinese. However, the abstracts will remain in English unless user clicks the translate button on an individual article card. For the full record page, both the title and abstract will be translated automatically, but only if user choose to translate abstract.
We plan to extend translated search to other UI language, thus allowing one-box multilingual capabilities.
Autocorrection feature (Handling Typos & Misspellings)
Even if you type something incorrectly, the system helps by:
- Autocorrecting mistakes if not results (showing results for ‘Animale’ instead of ‘Animal’).
- For misspelt words e.g. Dogg, it will show results and ask “Did you mean?” ‘Dog’ on summary page if input is unclear.
Example:
You type Physic → It will return results for "Physics"
You type Crispr gene editng → It suggests "Did you mean? Crispr gene editing on results page
- Personalized, Context-Aware Typeaheads
Intelligent typeahead help speeds up searches by predicting what user need, reduces effort by personalizing suggestions based on your previous searches for signed-in users and helps discover related research by suggesting relevant topics.
As users start typing, intelligent typeaheads provide suggestions based on:
- Popular Searches - Takes Web of Science history into consideration and when users sign in, the experience becomes even more personalized and context-aware
- Topics – Suggests topics relating to user query
- Author Names – Matches authors in Last Name, First Name format
- Affiliations – Suggests universities, institutions and research organizations.
- Journal Names – Suggests known journals as you type.
If you often search for "machine learning," you might see suggestions like "Machine Learning in Healthcare" or "Deep Learning Models." If you type "Smith, J," the system may prioritize "Smith, John (MIT)" based on your history.
Dual-Tab Interface – Search results
When you search using One-Box, results are neatly displayed on two separate tabs:
Documents Tab: Display document search results.
Researchers Tab: Display researchers, number of documents published, affiliation etc. and users can view their profiles.
| User Query | Document Tab (Search results) | Researcher Tab (Search results) |
| Author name(s) | Search results: document(s) published by the author |
Search results: authors that match the name (unclaimed and claimed)
|
| Topic |
Search results: document(s) on that topic (covid)
|
Search results: authors who write on that topic based on number of publication and citations. |
| Institution, Affiliation |
Search results: document(s) written by the address listed on the paper
|
Search results: authors who currently work at that institution or who have worked at institution
|
| Journal/Publication title |
Search results: publication(s) published in that journal
|
Search results: authors have published in that journal
|
| Article title |
Search results: document(s) that match the title
|
Search results: author(s) who wrote that publication(s)
|
| Author identifier |
Search results: document(s) written by that author(s) that match the identifier
|
Search results: author(s) that match the identifier
|
| Document identifier |
Search results: exact document specified
|
Search results: Authors that wrote the article
|
| Multiple fields (Author name and pub year; Author name affiliation or topic) |
Results come back with document(s) published by the author and affiliation (address listed) or topic
|
Search results: author profiles that match the name
|
- Document Tab (search results)
Boolean and Semantic Search Results
Once user search is processed, we combine two approaches:
- Boolean Search results (Exact Match & Text Matching, All fields)
- Users don’t need to use Boolean syntax, the system automatically structures the search correctly in the background and return Boolean search results for more accurate results.
- Entity recognition in One-Box (e.g., journal names, authors, DOIs, publication year, affiliation, article title).
- Finds results that exactly match user query and ensures precision
Example:
Climate change effects on agriculture
The system converts it into:
(climate change OR effects OR agriculture)
And returns only papers that contain these words only.
- Semantic Search (AI-Powered Meaning Matching)
- Uses AI & NLP to find conceptually similar papers.
- Expands searches to include synonyms & related topics.
- Helps when searches are broad or unclear.
Example:
Impact of global warming on food security
Semantic search might return:
“Climate Change Effects on Crop Yield”
“Extreme Weather and Agricultural Productivity”
“Global warming, food insecurity and crop yield”
How We Blend Boolean and Sematic Search
Boolean search provides precision.
Semantic search brings context-aware flexibility.
The system ranks and blends results for the best experience.
Example Hybrid Search Results for:
Impact of global warming on food security
1️) Paper titled " Food Security: The Challenge of Feeding 9 Billion People" (Boolean match).
2️) Paper titled “The Role of Digital Agriculture in Mitigating Climate Change and Ensuring Food Security: An Overview” (Semantic match).
Our ranking algorithm blends both sets of results, presenting you with the most relevant documents and researcher profiles on separate tabs.
- Researcher Tab (search results)
Author name search: In the researcher tab, we are replicating the functionalities of the Advanced Search page's Researcher Search. For name searches, use the format "Last name, First name" or select a suggestion from the one-box. If you enter "First name, Last name," it will be treated as a 'topic' rather than an 'author name' due to system constraints.
Author Topic Search: For searches specifically related to an author topic, we leverage the topic model to return top 100 authors relevant to that topic based on number of documents published and citation count.
For all other entity-based searches, the system returns authors based on their publications (e.g., those who have published in a particular journal or are affiliated with a specific institution).
Example:
User enter search term “AI ethics.”
We will not only return documents on AI ethics in the document tab but also, in a separate tab, displays a list of leading researchers (top 100 authors) working on AI ethics, not ranked but authors most publications and citations on AI ethics. For other types of searches, you we display authors associated with the relevant journals or institutions or user query.