xinmi/internal/application/service/chat_pipeline/extract_entity_price.go

package chatpipeline

import (
	"fmt"
	"regexp"
	"strings"

	"github.com/Tencent/WeKnora/internal/types"
)

var graphPriceLabelPattern = regexp.MustCompile(`(?:终端)?零售(?:价格|价)|批发(?:价格|价)|促销(?:价格|价)|优惠(?:价格|价)|活动(?:价格|价)|会员(?:价格|价)|建议零售价|销售价格|售价|单价|原价|现价|价格`)
var graphPriceAmountPattern = regexp.MustCompile(`^(人民币|[¥￥$])?([0-9]+(?:\.[0-9]+)?)(美元|人民币|元|角|分|盒|箱|瓶|套|支|粒|片|件|袋|kg|mg|ml|g)?(/(?:[0-9一二两俩三四五六七八九十百千]+)?(?:盒|箱|瓶|套|支|粒|片|件|袋|kg|mg|ml|g))?`)
var graphPriceHeaderUnitPattern = regexp.MustCompile(`^[（(](元|美元|元/(?:盒|箱|瓶|套|支|件))[）)]$`)

type graphPriceValue struct {
	number string
	unit   string
}

type graphPriceClaim struct {
	label       string
	values      []graphPriceValue
	prefix      string
	hasQuantity bool
}

func graphPriceValues(value, headerUnit string) []graphPriceValue {
	value = strings.TrimLeft(compactGraphText(value), "：:为是")
	var values []graphPriceValue
	for value != "" {
		match := graphPriceAmountPattern.FindStringSubmatch(value)
		if match == nil {
			// Do not accept a parsed prefix while silently ignoring a changed
			// package quantity or another amount later in the same price claim.
			if strings.HasPrefix(value, "/") || graphQuantityPattern.MatchString(value) {
				return nil
			}
			break
		}
		unit := match[3] + match[4]
		if match[1] != "" {
			currency := "元"
			if match[1] == "$" {
				currency = "美元"
			}
			if unit == "" || strings.HasPrefix(unit, "/") {
				unit = currency + unit
			}
		}
		if unit == "" {
			unit = headerUnit
		}
		values = append(values, graphPriceValue{number: match[2], unit: unit})
		value = strings.TrimLeft(value[len(match[0]):], "，,、")
	}
	return values
}

func graphPriceClaims(value string) []graphPriceClaim {
	value = compactGraphText(value)
	labels := graphPriceLabelPattern.FindAllStringIndex(value, -1)
	var claims []graphPriceClaim
	for i, span := range labels {
		end := len(value)
		if i+1 < len(labels) {
			end = labels[i+1][0]
		}
		prefix := value[:span[0]]
		if index := strings.LastIndexAny(prefix, "，,；;。|：:"); index >= 0 {
			prefix = strings.TrimLeft(prefix[index:], "，,；;。|：:")
		}
		claims = append(claims, graphPriceClaim{
			label:       graphPriceLabel(value[span[0]:span[1]]),
			values:      graphPriceValues(value[span[1]:end], ""),
			prefix:      prefix,
			hasQuantity: graphQuantityPattern.MatchString(value[span[1]:end]),
		})
	}
	return claims
}

func graphPriceLabel(value string) string {
	return strings.TrimPrefix(strings.ReplaceAll(value, "价格", "价"), "终端")
}

func graphPriceValuesMatch(want, available []graphPriceValue) bool {
	if len(want) == 0 || len(want) > len(available) {
		return false
	}
	for start := 0; start+len(want) <= len(available); start++ {
		matches := true
		for i, value := range want {
			actual := available[start+i]
			if value.number != actual.number || (value.unit != "" && value.unit != actual.unit) {
				matches = false
				break
			}
		}
		if matches {
			return true
		}
	}
	return false
}

func graphPricePrefixMatches(prefix, name string, inherited bool) bool {
	prefix = strings.TrimSuffix(compactGraphText(prefix), "的")
	if prefix == compactGraphText(name) {
		return true
	}
	return inherited && (prefix == "" || prefix == "本品" || prefix == "该产品" || prefix == "其")
}

func graphPriceSentenceSupports(sentence, name string, inherited bool, want graphPriceClaim) bool {
	for _, claim := range graphPriceClaims(sentence) {
		if !graphPricePrefixMatches(claim.prefix, name, inherited) {
			inherited = false
			continue
		}
		inherited = true
		if claim.label == want.label && graphPriceValuesMatch(want.values, claim.values) {
			return true
		}
	}
	return false
}

func graphPriceTableCells(line string) []string {
	line = strings.TrimSpace(line)
	if !strings.HasPrefix(line, "|") {
		return nil
	}
	return strings.Split(strings.TrimSuffix(strings.TrimPrefix(line, "|"), "|"), "|")
}

func graphPriceProductHeader(value string) bool {
	switch compactGraphText(value) {
	case "产品名称", "商品名称", "品名", "产品":
		return true
	default:
		return false
	}
}

func graphPriceTableRowMatches(cells, headers []string, name string) bool {
	name = compactGraphText(name)
	if name == "" {
		return false
	}
	if len(headers) == 0 {
		for _, cell := range cells {
			if compactGraphText(cell) == name {
				return true
			}
		}
		return false
	}
	if len(cells) > len(headers) {
		return false
	}
	productColumn, brandColumn := -1, -1
	for i, header := range headers {
		if graphPriceProductHeader(header) {
			if productColumn >= 0 {
				return false
			}
			productColumn = i
		}
		if header = compactGraphText(header); header == "品牌" || header == "品牌名称" {
			if brandColumn >= 0 {
				return false
			}
			brandColumn = i
		}
	}
	if productColumn < 0 || productColumn >= len(cells) {
		return false
	}
	product := compactGraphText(cells[productColumn])
	if product == "" {
		return false
	}
	if product == name {
		return true
	}
	if brandColumn < 0 || brandColumn >= len(cells) {
		return false
	}
	brand := compactGraphText(cells[brandColumn])
	return brand != "" && brand+product == name
}

func graphPriceAttributeSupported(content, name string, want graphPriceClaim) bool {
	var headers []string
	headingMatches := false
	for _, line := range strings.Split(content, "\n") {
		if cells := graphPriceTableCells(line); cells != nil {
			headingMatches = false
			isHeader := false
			for _, cell := range cells {
				if graphPriceProductHeader(cell) {
					isHeader = true
				}
			}
			if isHeader {
				headers = cells
				continue
			}
			// A short row may end midway through its final cell. Only closed
			// cells have a known boundary; never validate a truncated price.
			if len(cells) < len(headers) && !strings.HasSuffix(strings.TrimSpace(line), "|") {
				cells = cells[:len(cells)-1]
			}
			if !graphPriceTableRowMatches(cells, headers, name) {
				continue
			}
			for i, cell := range cells {
				if graphPriceSentenceSupports(cell, name, true, want) {
					return true
				}
				// A chunk may end before the row's final cells; only existing
				// leading cells can inherit their corresponding explicit header.
				if i >= len(headers) {
					continue
				}
				header := compactGraphText(headers[i])
				span := graphPriceLabelPattern.FindStringIndex(header)
				if span == nil || span[0] != 0 || graphPriceLabel(header[:span[1]]) != want.label {
					continue
				}
				unit := ""
				if suffix := header[span[1]:]; suffix != "" {
					match := graphPriceHeaderUnitPattern.FindStringSubmatch(suffix)
					if match == nil {
						continue
					}
					unit = match[1]
				}
				if graphPriceValuesMatch(want.values, graphPriceValues(cell, unit)) {
					return true
				}
			}
			continue
		}
		headers = nil
		trimmed := strings.TrimSpace(line)
		if strings.HasPrefix(trimmed, "#") || graphProductNameMarkerPattern.MatchString(trimmed) {
			heading := strings.TrimLeft(trimmed, "# ")
			heading = strings.TrimPrefix(graphProductIdentity(heading), "产品名称")
			headingMatches = heading == graphProductIdentity(name)
			continue
		}
		for _, sentence := range strings.FieldsFunc(line, func(r rune) bool { return strings.ContainsRune("。！？!?；;", r) }) {
			if graphPriceSentenceSupports(sentence, name, headingMatches, want) {
				return true
			}
		}
	}
	return false
}

func validateGraphPriceAttributes(graph *types.GraphData, content string) error {
	for i, node := range graph.Node {
		for j, attribute := range node.Attributes {
			for _, claim := range graphPriceClaims(attribute) {
				if !claim.hasQuantity {
					continue
				}
				if !graphPriceAttributeSupported(content, node.Name, claim) {
					return fmt.Errorf("%w: entity %d attribute %d has an unsupported price label/value (%s); require an explicit label and matching amount/unit in this product's sentence or table column, never infer unheaded prices", ErrGraphNeedsReview, i+1, j+1, claim.label)
				}
			}
		}
	}
	return nil
}