Multilingual Search & Research Interpreter

Break the language barrier in global research. Surface global insights with AI by automating query translation, Google search, and LLM summarization across 32 languages.

The Problem: Information Silos in a Multilingual World

The internet contains a wealth of knowledge, but the vast majority of it is invisible to monolingual searchers. Consider:

• Region-exclusive content: Local news, academic publications, government reports, and expert analyses are often published only in the native language of their origin.
• Search engine limitations: Google defaults to returning results in the language of your query, effectively hiding relevant content published in other languages.
• The researcher's dilemma: Experienced researchers know that translating their search queries into multiple languages yields dramatically different—and often more valuable—results. However, this process is time-consuming: translate the query, search, translate each result back, evaluate relevance, and synthesize findings.

This actor automates the entire multilingual research workflow, transforming what would take hours of manual effort into a streamlined, one-click operation.

How It Works

The Multilingual Research Synthesis Actor functions as an intelligent research interpreter:

1. Query Expansion — Automatically translates your search phrase into multiple target languages using AI-powered translation
2. Parallel Search — Executes Google searches in each language, surfacing region-specific content
3. Content Extraction — Crawls landing pages to retrieve full article content
4. Back-Translation — Translates all discovered content back to your source language
5. Intelligent Summarization — Uses LLM to generate concise summaries with relevance scoring
6. Deduplication — Removes near-duplicate results to eliminate redundancy
7. Export — Delivers a clean, unified research report in your preferred format

Use Cases

Scenario	Example
Market Research	Discover how a product is discussed in Japanese tech blogs vs. German industry publications
Academic Research	Find primary sources on historical events published in their country of origin
Competitive Intelligence	Monitor competitor coverage in local markets worldwide
Journalism	Access firsthand accounts and local reporting on global events
Policy Analysis	Compare how different countries report on regulatory changes
Medical Research	Surface clinical findings published in non-English journals

Key Features

• 32 Languages Supported — From English and Spanish to Japanese, Arabic, and beyond
• Auto-Detection — Automatically detects your query's language or lets you specify it
• 6 LLM Providers — Choose from OpenAI, Anthropic, Google, Groq, OpenRouter, or Perplexity
• Relevance Scoring — Each result includes a 0-100 relevance score relative to your original query
• Multiple Export Formats — CSV, XLSX, or TXT reports ready for further analysis

---

Input Schema

{
    "searchPhrase": "artificial intelligence in healthcare",
    "sourceLanguage": "auto",
    "targetLanguages": ["fr", "de", "es", "ja", "zh"],
    "maxResultsPerLanguage": 5,
    "recencyFilter": "month",
    "lingoApiKey": "YOUR_LINGO_DEV_API_KEY",
    "llmProvider": "openai",
    "llmApiKey": "YOUR_LLM_API_KEY",
    "outputFormat": "csv"
}

Input Fields

Field	Type	Required	Description
searchPhrase	string	Yes	The search query to research
sourceLanguage	string	No	Language of input/output (auto-detected by default)
targetLanguages	array	Yes	Languages to search in (min 1)
maxResultsPerLanguage	integer	Yes	Max results per language (1-10)
recencyFilter	enum	No	Filter: week, month, year, none
lingoApiKey	string	Yes	Your Lingo.dev API key for translation
llmProvider	string	No	LLM provider: openai, anthropic, google, groq, openrouter, perplexity
llmApiKey	string	Yes	Your API key for the selected LLM provider
outputFormat	enum	Yes	Export format: csv, xlsx, txt

Output Schema

Each result record contains:

{
    "targetLanguage": "fr",
    "summary": "Summary in your source language...",
    "link": "https://example.com/article",
    "pageDate": "2024-01-15",
    "relevanceScore": 85
}

Architecture

Component	Technology
Translation	Lingo.dev AI-powered translation
Search	Apify Google Search Actor
Crawl	Crawlee CheerioCrawler (landing pages only)
LLM	Multi-provider support (OpenAI, Anthropic, Google, Groq, OpenRouter, Perplexity)
Export	CSV, XLSX, TXT formats

Design Principles

• Landing pages only — No internal link following to respect site boundaries
• Per-language isolation — One language failing doesn't stop the entire run
• LLM guardrails — Strict prompts prevent hallucination and ensure accuracy
• Token limits — Hard limits on LLM usage for cost predictability
• Transparency — Skipped pages are logged with reasons

Local Development

# Install dependencies
npm install

# Set test input in storage/key_value_stores/default/INPUT.json

# Run locally
npx apify run

# Check output in storage/datasets/default/

---

Built for researchers who refuse to let language be a barrier to knowledge.