如何在Watson Conversation（Bluemix）中动态映射输入与实体

Hey there! Let's work through how to build that bird entity mapping for your chatbot—whether users ask about a single species like sparrows or a group like peacocks and parrots, here's a solid approach to handle dynamic inputs:

First, you need a centralized library to store your bird entities, including any common aliases users might use. This makes it easy to map varied user inputs to standard entity names. For example:

# A flexible dictionary to hold bird entities and their aliases
bird_entity_library = {
    "麻雀": {"aliases": ["家雀", "树麻雀"]},
    "孔雀": {"aliases": ["绿孔雀", "蓝孔雀"]},
    "鹦鹉": {"aliases": ["鹦哥", "金刚鹦鹉", "虎皮鹦鹉"]}
}

Pro tip: Include regional nicknames or common misspellings here to cover more user variations.

Next, you need to pull bird names from the user's query and map them back to your standard entities. Here's a practical Python implementation that handles both single and multiple bird inputs:

import re

def extract_and_map_birds(user_input):
    # Combine all bird names and aliases into a single list for matching
    all_bird_terms = []
    for bird, details in bird_entity_library.items():
        all_bird_terms.append(bird)
        all_bird_terms.extend(details["aliases"])
    
    # Use regex to find all matching bird terms in the input (case-insensitive)
    pattern = re.compile(r'\b(' + '|'.join(re.escape(term) for term in all_bird_terms) + r')\b', re.IGNORECASE)
    matched_terms = pattern.findall(user_input)
    
    # Map matched terms back to standard entity names
    mapped_birds = []
    for term in matched_terms:
        term_lower = term.lower()
        for bird, details in bird_entity_library.items():
            if term_lower == bird.lower() or term_lower in [alias.lower() for alias in details["aliases"]]:
                mapped_birds.append(bird)
                break
    
    # Remove duplicates in case the user mentions a bird multiple times
    return list(set(mapped_birds))

# Test with your example queries
print(extract_and_map_birds("我想了解麻雀、孔雀、鹦鹉这类鸟类"))  # Output: ['麻雀', '孔雀', '鹦鹉']
print(extract_and_map_birds("我想了解家雀这种鸟类"))  # Output: ['麻雀']

Users might not always use exact bird names. For example, if someone types "我想了解会说话的鸟", you'll need to:

• Pair entity extraction with intent recognition to understand the user's goal
• Use fuzzy matching (libraries like fuzzywuzzy can help) to match similar terms to your entity library
• Fall back to a confirmation prompt: "Did you mean 鹦鹉?" if the input is unclear

Since users can ask about any bird, don't hardcode your entity library. Instead, load it from a dynamic source like a CSV or database so you can easily add new species without rewriting code:

import csv

def load_bird_library_from_csv(csv_file_path):
    bird_library = {}
    with open(csv_file_path, 'r', encoding='utf-8') as file:
        reader = csv.DictReader(file)
        for row in reader:
            bird_name = row["standard_name"]
            aliases = row["aliases"].split(',') if row["aliases"] else []
            bird_library[bird_name] = {"aliases": aliases}
    return bird_library

# Usage: bird_entity_library = load_bird_library_from_csv("bird_entities.csv")

This setup will let your chatbot reliably map both single and multiple dynamic bird inputs to their corresponding entities, no matter how users phrase their questions.

内容的提问来源于stack exchange，提问作者user1352892