StaffPerformance/wispro.sp.utility/CosineSimilarity.cs

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;

namespace wispro.sp.utility
{
    public class CosineSimilarity
    {
        public static double Calculate(string s1, string s2)
        {
            // 将字符串转换为词频向量
            var vector1 = GetTermFrequencyVector(s1);
            var vector2 = GetTermFrequencyVector(s2);

            // 计算余弦相似度
            double dotProduct = 0;
            double norm1 = 0;
            double norm2 = 0;

            foreach (var term in vector1.Keys.Union(vector2.Keys))
            {
                double v1 = vector1.ContainsKey(term) ? vector1[term] : 0;
                double v2 = vector2.ContainsKey(term) ? vector2[term] : 0;

                dotProduct += v1 * v2;
                norm1 += v1 * v1;
                norm2 += v2 * v2;
            }

            return dotProduct / (Math.Sqrt(norm1) * Math.Sqrt(norm2));
        }

        private static Dictionary<string, double> Normalize(Dictionary<string, double> vector)
        {
            double length = 0;
            foreach (var key in vector.Keys)
            {
                length += vector.GetValueOrDefault(key) * vector.GetValueOrDefault(key);
            }
            length = Math.Sqrt(length);

            Dictionary<string,double> result = new Dictionary<string, double>();
            foreach (var key in vector.Keys)
            {
                result.Add(key, vector.GetValueOrDefault(key)/length);
            }
            
            return result;
        }


        private static Dictionary<string, double> GetTermFrequencyVector(string text)
        {
            var terms = new Jieba_Segmenter().Cut(text.Replace("\r\n","").Replace("\r","").Replace("\n", ""));
            var vector = new Dictionary<string, double>();

            foreach (var term in terms)
            {
                if (!vector.ContainsKey(term))
                    vector[term] = 0;
                vector[term]++;
            }

            return Normalize(vector);
        }
    }
}