func parseCaptions(text: String) -> String {
        let textRange = NSRange(location: 0, length: text.utf16.count) //NSRange, based on NSString use UTF16 for counting, while Swift.String use UTF8 by default, so `text.count` might be wrong
        let regex = try! NSRegularExpression(pattern: "<text[^>]+>([^<]+)<\\/text>")
        let matches = regex.matches(in: text, range: textRange)

        var result: String = text
        for match in matches.reversed() {
            let wholeNSRange = match.range
            let wholeRange = Range(wholeNSRange, in: result)!
            let textNSRange = match.range(at: 1)
            let textRange = Range(textNSRange, in: result)!
            var string = String(result[textRange])
            string = string.replacingOccurrences(of: "\n", with: " ")
            string = string.replacingOccurrences(of: "&#39;", with: "'")
            string = string.replacingOccurrences(of: "&amp;quot;", with: "\"")
            string.append("\n")
            result = result.replacingCharacters(in: wholeRange, with: string)
        }
        return result
    }